{
  "source": "https://modelverdict.ai",
  "license": "CC BY 4.0",
  "run_id": "20260928T090853Z-cb9757",
  "run_date": "2026-09-28",
  "dataset_version": "0.1.0",
  "leaderboards": [
    {
      "task": "invoices",
      "lang": "cs",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 9.4119,
          "latency_ms_p50": 3608,
          "latency_ms_p90": 4354,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 4.2591,
          "latency_ms_p50": 3450,
          "latency_ms_p90": 5273,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9933,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.4962,
          "latency_ms_p50": 4960,
          "latency_ms_p90": 7713,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9933,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9933
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 10.1173,
          "latency_ms_p50": 3855,
          "latency_ms_p90": 5418,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9933,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9933,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 1.0313,
          "latency_ms_p50": 1802,
          "latency_ms_p90": 2186,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9533,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9867,
            "bank_account": 0.9933
          },
          "errors": {
            "hallucinated": 7,
            "wrong_value": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9928,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.5051,
          "latency_ms_p50": 3173,
          "latency_ms_p90": 4223,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9533,
            "supplier_id": 0.9867,
            "vat_id": 0.98,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 0.9933,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9933
          },
          "errors": {
            "wrong_value": 12,
            "lost_diacritics": 1,
            "swapped_dates": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.961,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 0.2118,
          "latency_ms_p50": 10427,
          "latency_ms_p90": 14675,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 0.9333,
            "vat_id": 0.9667,
            "invoice_number": 1,
            "payment_reference": 0.96,
            "issue_date": 0.96,
            "tax_point_date": 0.9533,
            "due_date": 0.94,
            "net_by_rate": 0.98,
            "vat_by_rate": 0.9733,
            "total": 0.9933,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 66,
            "missing_field": 8,
            "swapped_dates": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9364,
          "doc_accuracy": 0.56,
          "usd_per_1000": 0.4373,
          "latency_ms_p50": 1999,
          "latency_ms_p90": 2851,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9133,
            "supplier_id": 0.9267,
            "vat_id": 0.94,
            "invoice_number": 1,
            "payment_reference": 0.9733,
            "issue_date": 0.98,
            "tax_point_date": 0.76,
            "due_date": 0.9733,
            "net_by_rate": 0.9,
            "vat_by_rate": 0.9467,
            "total": 0.9867,
            "currency": 0.9733,
            "bank_account": 0.9
          },
          "errors": {
            "wrong_value": 59,
            "hallucinated": 57,
            "lost_diacritics": 4,
            "swapped_dates": 4
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9349,
          "doc_accuracy": 0.6333,
          "usd_per_1000": 4.6878,
          "latency_ms_p50": 2626,
          "latency_ms_p90": 3553,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9,
            "supplier_id": 0.8733,
            "vat_id": 0.8533,
            "invoice_number": 0.9933,
            "payment_reference": 0.9533,
            "issue_date": 0.8867,
            "tax_point_date": 0.8667,
            "due_date": 0.9067,
            "net_by_rate": 0.9933,
            "vat_by_rate": 0.9867,
            "total": 1,
            "currency": 1,
            "bank_account": 0.94
          },
          "errors": {
            "wrong_value": 99,
            "hallucinated": 18,
            "missing_field": 6,
            "lost_diacritics": 2,
            "swapped_dates": 2
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.7667,
          "usd_per_1000": 1.0185,
          "latency_ms_p50": 17027,
          "latency_ms_p90": 29813,
          "invalid_json_rate": 0.0333,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.94,
            "vat_id": 0.94,
            "invoice_number": 0.9467,
            "payment_reference": 0.94,
            "issue_date": 0.9467,
            "tax_point_date": 0.9467,
            "due_date": 0.9467,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.8067,
            "total": 0.94,
            "currency": 0.94,
            "bank_account": 0.9067
          },
          "errors": {
            "invalid_json": 65,
            "missing_field": 41,
            "wrong_value": 36,
            "wrong_rate": 16,
            "hallucinated": 1,
            "net_vs_total": 1
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "cs",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.8743,
          "latency_ms_p50": 2686,
          "latency_ms_p90": 3732,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.4951,
          "latency_ms_p50": 6034,
          "latency_ms_p90": 8902,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.9135,
          "latency_ms_p50": 1069,
          "latency_ms_p90": 1209,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.4604,
          "latency_ms_p50": 2842,
          "latency_ms_p90": 4044,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.2113,
          "latency_ms_p50": 2559,
          "latency_ms_p90": 3022,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.0152,
          "latency_ms_p50": 2820,
          "latency_ms_p90": 4174,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9983,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 2.9651,
          "latency_ms_p50": 1473,
          "latency_ms_p90": 1683,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.2422,
          "latency_ms_p50": 9030,
          "latency_ms_p90": 11492,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "missing_field": 2,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9675,
          "doc_accuracy": 0.7556,
          "usd_per_1000": 0.2551,
          "latency_ms_p50": 1426,
          "latency_ms_p90": 1988,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 0.8222,
            "due_date": 1,
            "net_by_rate": 0.9111,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9556
          },
          "errors": {
            "hallucinated": 12,
            "wrong_value": 7
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9197,
          "doc_accuracy": 0.5333,
          "usd_per_1000": 0.7421,
          "latency_ms_p50": 19899,
          "latency_ms_p90": 33538,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.5333,
            "vat_by_rate": 0.5556,
            "total": 0.9778,
            "currency": 0.9778,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 28,
            "wrong_rate": 16,
            "missing_field": 2,
            "net_vs_total": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "cs",
      "level": "clean",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 13.1965,
          "latency_ms_p50": 3788,
          "latency_ms_p90": 4090,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.4873,
          "latency_ms_p50": 4448,
          "latency_ms_p90": 7183,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.6141,
          "latency_ms_p50": 4013,
          "latency_ms_p90": 5479,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 14.5657,
          "latency_ms_p50": 3857,
          "latency_ms_p90": 5248,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9966,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.7289,
          "latency_ms_p50": 3625,
          "latency_ms_p90": 4560,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 0.9778,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1,
            "swapped_dates": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 1.0712,
          "latency_ms_p50": 2030,
          "latency_ms_p90": 2352,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 3
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9538,
          "doc_accuracy": 0.6222,
          "usd_per_1000": 0.1974,
          "latency_ms_p50": 10967,
          "latency_ms_p90": 15239,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.8889,
            "vat_id": 0.9556,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9556,
            "tax_point_date": 0.9556,
            "due_date": 0.9778,
            "net_by_rate": 0.9556,
            "vat_by_rate": 0.9778,
            "total": 0.9778,
            "currency": 1,
            "bank_account": 0.8889
          },
          "errors": {
            "wrong_value": 27
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9265,
          "doc_accuracy": 0.8444,
          "usd_per_1000": 1.1092,
          "latency_ms_p50": 15940,
          "latency_ms_p90": 25468,
          "invalid_json_rate": 0.0444,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.8444
          },
          "errors": {
            "invalid_json": 26,
            "missing_field": 13,
            "wrong_value": 4
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9231,
          "doc_accuracy": 0.4444,
          "usd_per_1000": 0.5117,
          "latency_ms_p50": 1877,
          "latency_ms_p90": 2277,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8889,
            "supplier_id": 0.9111,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 0.9556,
            "issue_date": 0.9778,
            "tax_point_date": 0.6667,
            "due_date": 1,
            "net_by_rate": 0.8889,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 0.9778,
            "bank_account": 0.8667
          },
          "errors": {
            "hallucinated": 25,
            "wrong_value": 18,
            "lost_diacritics": 2
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9214,
          "doc_accuracy": 0.6,
          "usd_per_1000": 5.372,
          "latency_ms_p50": 2486,
          "latency_ms_p90": 3741,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8889,
            "supplier_id": 0.8667,
            "vat_id": 0.8444,
            "invoice_number": 0.9778,
            "payment_reference": 0.9333,
            "issue_date": 0.8667,
            "tax_point_date": 0.7778,
            "due_date": 0.8667,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9556
          },
          "errors": {
            "wrong_value": 36,
            "hallucinated": 7,
            "missing_field": 2,
            "swapped_dates": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "cs",
      "level": "scan",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 10.4868,
          "latency_ms_p50": 3695,
          "latency_ms_p90": 4464,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9983,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 4.6137,
          "latency_ms_p50": 3547,
          "latency_ms_p90": 6085,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9983,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 11.8908,
          "latency_ms_p50": 4233,
          "latency_ms_p90": 5444,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9778,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9966,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.4929,
          "latency_ms_p50": 4783,
          "latency_ms_p90": 6733,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 1.0814,
          "latency_ms_p50": 1828,
          "latency_ms_p90": 2186,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9556,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 3,
            "hallucinated": 3
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9829,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.5858,
          "latency_ms_p50": 3361,
          "latency_ms_p90": 4223,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9111,
            "supplier_id": 0.9556,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 9,
            "lost_diacritics": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9487,
          "doc_accuracy": 0.5556,
          "usd_per_1000": 0.1985,
          "latency_ms_p50": 11919,
          "latency_ms_p90": 15914,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9111,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 0.9556,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9111,
            "net_by_rate": 1,
            "vat_by_rate": 0.9556,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8222
          },
          "errors": {
            "wrong_value": 25,
            "missing_field": 4,
            "swapped_dates": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9214,
          "doc_accuracy": 0.4667,
          "usd_per_1000": 0.5124,
          "latency_ms_p50": 2559,
          "latency_ms_p90": 2943,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.8889,
            "vat_id": 0.9111,
            "invoice_number": 1,
            "payment_reference": 0.9556,
            "issue_date": 0.9778,
            "tax_point_date": 0.7556,
            "due_date": 0.9556,
            "net_by_rate": 0.8889,
            "vat_by_rate": 0.9111,
            "total": 0.9778,
            "currency": 0.9333,
            "bank_account": 0.8889
          },
          "errors": {
            "wrong_value": 26,
            "hallucinated": 16,
            "swapped_dates": 3,
            "lost_diacritics": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9094,
          "doc_accuracy": 0.8889,
          "usd_per_1000": 1.1724,
          "latency_ms_p50": 15379,
          "latency_ms_p90": 32846,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.9111,
            "supplier_id": 0.9111,
            "vat_id": 0.9111,
            "invoice_number": 0.9111,
            "payment_reference": 0.9111,
            "issue_date": 0.9111,
            "tax_point_date": 0.9111,
            "due_date": 0.9111,
            "net_by_rate": 0.9111,
            "vat_by_rate": 0.9111,
            "total": 0.9111,
            "currency": 0.9111,
            "bank_account": 0.8889
          },
          "errors": {
            "invalid_json": 39,
            "missing_field": 13,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.8906,
          "doc_accuracy": 0.3556,
          "usd_per_1000": 5.4072,
          "latency_ms_p50": 2905,
          "latency_ms_p90": 3837,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8222,
            "supplier_id": 0.7111,
            "vat_id": 0.6667,
            "invoice_number": 1,
            "payment_reference": 0.9111,
            "issue_date": 0.8444,
            "tax_point_date": 0.8444,
            "due_date": 0.9333,
            "net_by_rate": 0.9778,
            "vat_by_rate": 0.9778,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8889
          },
          "errors": {
            "wrong_value": 49,
            "hallucinated": 11,
            "missing_field": 3,
            "lost_diacritics": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "cs",
      "level": "photo",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 8.4467,
          "latency_ms_p50": 3857,
          "latency_ms_p90": 4384,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.5362,
          "latency_ms_p50": 5361,
          "latency_ms_p90": 6488,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.5263,
          "latency_ms_p50": 3367,
          "latency_ms_p90": 4783,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 1.1148,
          "latency_ms_p50": 1821,
          "latency_ms_p90": 2119,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 9.7578,
          "latency_ms_p50": 4944,
          "latency_ms_p90": 6437,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.4723,
          "latency_ms_p50": 3293,
          "latency_ms_p90": 3794,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9282,
          "doc_accuracy": 0.6,
          "usd_per_1000": 0.5354,
          "latency_ms_p50": 2851,
          "latency_ms_p90": 3432,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8667,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9333,
            "tax_point_date": 0.8667,
            "due_date": 0.8667,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 8,
            "hallucinated": 4,
            "lost_diacritics": 1,
            "swapped_dates": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.3333,
          "usd_per_1000": 0.2037,
          "latency_ms_p50": 10965,
          "latency_ms_p90": 17598,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9333,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.8667,
            "due_date": 0.7333,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7333
          },
          "errors": {
            "wrong_value": 13,
            "missing_field": 2,
            "swapped_dates": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.5333,
          "usd_per_1000": 5.645,
          "latency_ms_p50": 2995,
          "latency_ms_p90": 3778,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.7333,
            "tax_point_date": 0.8,
            "due_date": 0.6667,
            "net_by_rate": 1,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 14,
            "lost_diacritics": 1,
            "swapped_dates": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9128,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 1.1136,
          "latency_ms_p50": 15434,
          "latency_ms_p90": 43358,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8667,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.8667,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.8667,
            "vat_by_rate": 0.8667,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.9333
          },
          "errors": {
            "missing_field": 13,
            "wrong_value": 3,
            "hallucinated": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "missing_field",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "de",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.03,
          "latency_ms_p50": 1530,
          "latency_ms_p90": 1909,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 10.652,
          "latency_ms_p50": 4897,
          "latency_ms_p90": 6986,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9933
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9954,
          "doc_accuracy": 0.94,
          "usd_per_1000": 0.5657,
          "latency_ms_p50": 6405,
          "latency_ms_p90": 9388,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9933,
            "supplier_id": 0.9867,
            "vat_id": 0.9933,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 0.9933,
            "due_date": 0.9933,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.98
          },
          "errors": {
            "wrong_value": 8,
            "swapped_dates": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9133,
          "usd_per_1000": 0.5446,
          "latency_ms_p50": 3701,
          "latency_ms_p90": 5093,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9867,
            "vat_id": 0.9933,
            "invoice_number": 0.9933,
            "payment_reference": 1,
            "issue_date": 0.9867,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.94
          },
          "errors": {
            "wrong_value": 13,
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 10.4175,
          "latency_ms_p50": 3751,
          "latency_ms_p90": 5762,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9933,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8933,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.98
          },
          "errors": {
            "missing_field": 17,
            "wrong_value": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9738,
          "doc_accuracy": 0.7467,
          "usd_per_1000": 4.7201,
          "latency_ms_p50": 2478,
          "latency_ms_p90": 3177,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9733,
            "supplier_id": 0.9867,
            "vat_id": 0.96,
            "invoice_number": 1,
            "payment_reference": 0.9267,
            "issue_date": 0.98,
            "tax_point_date": 0.9067,
            "due_date": 0.9733,
            "net_by_rate": 0.98,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9733
          },
          "errors": {
            "hallucinated": 26,
            "wrong_value": 18,
            "missing_field": 5,
            "swapped_dates": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9723,
          "doc_accuracy": 0.96,
          "usd_per_1000": 7.1537,
          "latency_ms_p50": 4732,
          "latency_ms_p90": 8147,
          "invalid_json_rate": 0.0267,
          "by_field": {
            "supplier_name": 0.9733,
            "supplier_id": 0.9733,
            "vat_id": 0.9667,
            "invoice_number": 0.9733,
            "payment_reference": 0.9733,
            "issue_date": 0.9733,
            "tax_point_date": 0.9733,
            "due_date": 0.9733,
            "net_by_rate": 0.9667,
            "vat_by_rate": 0.9733,
            "total": 0.9733,
            "currency": 0.9733,
            "bank_account": 0.9733
          },
          "errors": {
            "invalid_json": 52,
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9487,
          "doc_accuracy": 0.64,
          "usd_per_1000": 0.199,
          "latency_ms_p50": 7704,
          "latency_ms_p90": 13269,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9867,
            "supplier_id": 0.9933,
            "vat_id": 0.94,
            "invoice_number": 0.9933,
            "payment_reference": 0.9867,
            "issue_date": 0.8733,
            "tax_point_date": 0.92,
            "due_date": 0.9067,
            "net_by_rate": 0.9467,
            "vat_by_rate": 0.9733,
            "total": 0.9933,
            "currency": 0.9933,
            "bank_account": 0.8267
          },
          "errors": {
            "wrong_value": 90,
            "missing_field": 7,
            "swapped_dates": 3
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9462,
          "doc_accuracy": 0.5,
          "usd_per_1000": 0.4365,
          "latency_ms_p50": 1960,
          "latency_ms_p90": 2664,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9857,
            "supplier_id": 0.8,
            "vat_id": 0.9286,
            "invoice_number": 1,
            "payment_reference": 0.8714,
            "issue_date": 1,
            "tax_point_date": 0.8786,
            "due_date": 0.9857,
            "net_by_rate": 0.9429,
            "vat_by_rate": 0.9929,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9143
          },
          "errors": {
            "hallucinated": 64,
            "wrong_value": 22,
            "missing_field": 10,
            "swapped_dates": 2
          },
          "n_cases": 140,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9349,
          "doc_accuracy": 0.8733,
          "usd_per_1000": 1.1188,
          "latency_ms_p50": 19027,
          "latency_ms_p90": 33896,
          "invalid_json_rate": 0.06,
          "by_field": {
            "supplier_name": 0.94,
            "supplier_id": 0.94,
            "vat_id": 0.94,
            "invoice_number": 0.9267,
            "payment_reference": 0.8867,
            "issue_date": 0.94,
            "tax_point_date": 0.94,
            "due_date": 0.94,
            "net_by_rate": 0.94,
            "vat_by_rate": 0.94,
            "total": 0.94,
            "currency": 0.94,
            "bank_account": 0.94
          },
          "errors": {
            "invalid_json": 117,
            "hallucinated": 8,
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "de",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.492,
          "latency_ms_p50": 6669,
          "latency_ms_p90": 9667,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.9111,
          "latency_ms_p50": 1100,
          "latency_ms_p90": 1283,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.2175,
          "latency_ms_p50": 2983,
          "latency_ms_p90": 3682,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.8529,
          "latency_ms_p50": 3109,
          "latency_ms_p90": 4157,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.9868,
          "latency_ms_p50": 27816,
          "latency_ms_p90": 50076,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.227,
          "latency_ms_p50": 7019,
          "latency_ms_p90": 12184,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 2.9567,
          "latency_ms_p50": 1392,
          "latency_ms_p90": 1702,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9556,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 2,
            "hallucinated": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9932,
          "doc_accuracy": 0.9111,
          "usd_per_1000": 5.8922,
          "latency_ms_p50": 3272,
          "latency_ms_p90": 5883,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9778,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9556,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "missing_field": 3,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.978,
          "doc_accuracy": 0.7857,
          "usd_per_1000": 0.2471,
          "latency_ms_p50": 1326,
          "latency_ms_p90": 1709,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9762,
            "invoice_number": 1,
            "payment_reference": 0.9048,
            "issue_date": 1,
            "tax_point_date": 0.9762,
            "due_date": 1,
            "net_by_rate": 0.8571,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 10,
            "missing_field": 1,
            "wrong_value": 1
          },
          "n_cases": 42,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9778,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 6.2407,
          "latency_ms_p50": 3872,
          "latency_ms_p90": 7363,
          "invalid_json_rate": 0.0222,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 0.9778,
            "payment_reference": 0.9778,
            "issue_date": 0.9778,
            "tax_point_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.9778,
            "vat_by_rate": 0.9778,
            "total": 0.9778,
            "currency": 0.9778,
            "bank_account": 0.9778
          },
          "errors": {
            "invalid_json": 13
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "de",
      "level": "clean",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.0733,
          "latency_ms_p50": 1727,
          "latency_ms_p90": 1932,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 15.2566,
          "latency_ms_p50": 5582,
          "latency_ms_p90": 8432,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9966,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.5534,
          "latency_ms_p50": 5725,
          "latency_ms_p90": 9363,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9966,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.7681,
          "latency_ms_p50": 4092,
          "latency_ms_p90": 5171,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9556
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 13.2731,
          "latency_ms_p50": 3761,
          "latency_ms_p90": 4940,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8889,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "missing_field": 5,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9658,
          "doc_accuracy": 0.6889,
          "usd_per_1000": 5.4305,
          "latency_ms_p50": 2503,
          "latency_ms_p90": 3311,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 1,
            "vat_id": 0.9111,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 0.9778,
            "tax_point_date": 0.8667,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "hallucinated": 11,
            "wrong_value": 8,
            "missing_field": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9521,
          "doc_accuracy": 0.9111,
          "usd_per_1000": 8.8069,
          "latency_ms_p50": 5108,
          "latency_ms_p90": 8285,
          "invalid_json_rate": 0.0444,
          "by_field": {
            "supplier_name": 0.9556,
            "supplier_id": 0.9556,
            "vat_id": 0.9333,
            "invoice_number": 0.9556,
            "payment_reference": 0.9556,
            "issue_date": 0.9556,
            "tax_point_date": 0.9556,
            "due_date": 0.9556,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9556,
            "total": 0.9556,
            "currency": 0.9556,
            "bank_account": 0.9556
          },
          "errors": {
            "invalid_json": 26,
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9385,
          "doc_accuracy": 0.5111,
          "usd_per_1000": 0.1859,
          "latency_ms_p50": 7433,
          "latency_ms_p90": 11387,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 1,
            "vat_id": 0.9556,
            "invoice_number": 0.9778,
            "payment_reference": 0.9556,
            "issue_date": 0.8444,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.9111,
            "vat_by_rate": 0.9556,
            "total": 0.9778,
            "currency": 1,
            "bank_account": 0.7778
          },
          "errors": {
            "wrong_value": 35,
            "missing_field": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9353,
          "doc_accuracy": 0.3636,
          "usd_per_1000": 0.5141,
          "latency_ms_p50": 1806,
          "latency_ms_p90": 2273,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.7045,
            "vat_id": 0.8864,
            "invoice_number": 1,
            "payment_reference": 0.8636,
            "issue_date": 1,
            "tax_point_date": 0.7955,
            "due_date": 0.9773,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9318
          },
          "errors": {
            "hallucinated": 26,
            "wrong_value": 6,
            "missing_field": 5
          },
          "n_cases": 44,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8838,
          "doc_accuracy": 0.8222,
          "usd_per_1000": 1.2649,
          "latency_ms_p50": 16696,
          "latency_ms_p90": 26372,
          "invalid_json_rate": 0.1111,
          "by_field": {
            "supplier_name": 0.8889,
            "supplier_id": 0.8889,
            "vat_id": 0.8889,
            "invoice_number": 0.8667,
            "payment_reference": 0.8444,
            "issue_date": 0.8889,
            "tax_point_date": 0.8889,
            "due_date": 0.8889,
            "net_by_rate": 0.8889,
            "vat_by_rate": 0.8889,
            "total": 0.8889,
            "currency": 0.8889,
            "bank_account": 0.8889
          },
          "errors": {
            "invalid_json": 65,
            "hallucinated": 2,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "de",
      "level": "scan",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.0799,
          "latency_ms_p50": 1606,
          "latency_ms_p90": 1929,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9983,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 13.1538,
          "latency_ms_p50": 5077,
          "latency_ms_p90": 7039,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.6076,
          "latency_ms_p50": 7201,
          "latency_ms_p90": 8817,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 0.9778,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 6
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9863,
          "doc_accuracy": 0.8222,
          "usd_per_1000": 11.8759,
          "latency_ms_p50": 3838,
          "latency_ms_p90": 5963,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8222,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 8
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9829,
          "doc_accuracy": 0.8222,
          "usd_per_1000": 0.6634,
          "latency_ms_p50": 4068,
          "latency_ms_p90": 5384,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9556,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9556,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8889
          },
          "errors": {
            "wrong_value": 8,
            "missing_field": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9778,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 6.6488,
          "latency_ms_p50": 4988,
          "latency_ms_p90": 9991,
          "invalid_json_rate": 0.0222,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 0.9778,
            "payment_reference": 0.9778,
            "issue_date": 0.9778,
            "tax_point_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.9778,
            "vat_by_rate": 0.9778,
            "total": 0.9778,
            "currency": 0.9778,
            "bank_account": 0.9778
          },
          "errors": {
            "invalid_json": 13
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9692,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 5.4545,
          "latency_ms_p50": 2767,
          "latency_ms_p90": 3225,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9556,
            "vat_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 0.9778,
            "tax_point_date": 0.9111,
            "due_date": 0.9556,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "hallucinated": 11,
            "wrong_value": 7
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9692,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 1.1487,
          "latency_ms_p50": 16648,
          "latency_ms_p90": 24926,
          "invalid_json_rate": 0.0222,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 0.9778,
            "invoice_number": 0.9778,
            "payment_reference": 0.8667,
            "issue_date": 0.9778,
            "tax_point_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.9778,
            "vat_by_rate": 0.9778,
            "total": 0.9778,
            "currency": 0.9778,
            "bank_account": 0.9778
          },
          "errors": {
            "invalid_json": 13,
            "hallucinated": 5
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9287,
          "doc_accuracy": 0.3659,
          "usd_per_1000": 0.5161,
          "latency_ms_p50": 2364,
          "latency_ms_p90": 2794,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9756,
            "supplier_id": 0.7073,
            "vat_id": 0.9024,
            "invoice_number": 1,
            "payment_reference": 0.8293,
            "issue_date": 1,
            "tax_point_date": 0.9024,
            "due_date": 1,
            "net_by_rate": 0.9512,
            "vat_by_rate": 0.9756,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8293
          },
          "errors": {
            "hallucinated": 22,
            "wrong_value": 11,
            "missing_field": 4,
            "swapped_dates": 1
          },
          "n_cases": 41,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.5111,
          "usd_per_1000": 0.1872,
          "latency_ms_p50": 8651,
          "latency_ms_p90": 13315,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9778,
            "vat_id": 0.9556,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.7778,
            "tax_point_date": 0.8444,
            "due_date": 0.8,
            "net_by_rate": 0.9111,
            "vat_by_rate": 0.9556,
            "total": 1,
            "currency": 0.9778,
            "bank_account": 0.7556
          },
          "errors": {
            "wrong_value": 42,
            "swapped_dates": 3,
            "missing_field": 3
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "de",
      "level": "photo",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.1068,
          "latency_ms_p50": 1609,
          "latency_ms_p90": 1823,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 6.448,
          "latency_ms_p50": 4808,
          "latency_ms_p90": 6024,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 9.7305,
          "latency_ms_p50": 5428,
          "latency_ms_p90": 6576,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9949,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.6975,
          "latency_ms_p50": 6384,
          "latency_ms_p90": 10996,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 0.9333,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "swapped_dates": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "swapped_dates",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 11.0515,
          "latency_ms_p50": 3820,
          "latency_ms_p90": 15338,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9333,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 1,
            "missing_field": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8,
          "usd_per_1000": 0.499,
          "latency_ms_p50": 3546,
          "latency_ms_p90": 5097,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 0.9333,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 3
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9487,
          "doc_accuracy": 0.6,
          "usd_per_1000": 5.6755,
          "latency_ms_p50": 2736,
          "latency_ms_p90": 3525,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.7333,
            "due_date": 0.8667,
            "net_by_rate": 0.9333,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 3,
            "hallucinated": 3,
            "swapped_dates": 2,
            "missing_field": 2
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9349,
          "doc_accuracy": 0.4615,
          "usd_per_1000": 0.5349,
          "latency_ms_p50": 2495,
          "latency_ms_p90": 2983,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9231,
            "supplier_id": 0.7692,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9231,
            "issue_date": 1,
            "tax_point_date": 0.7692,
            "due_date": 0.9231,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8462
          },
          "errors": {
            "hallucinated": 6,
            "wrong_value": 4,
            "swapped_dates": 1
          },
          "n_cases": 13,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9333,
          "doc_accuracy": 0.5333,
          "usd_per_1000": 0.1899,
          "latency_ms_p50": 9081,
          "latency_ms_p90": 13354,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8667,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.8667,
            "tax_point_date": 0.8667,
            "due_date": 0.8667,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.6667
          },
          "errors": {
            "wrong_value": 13
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.7897,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 0.987,
          "latency_ms_p50": 15594,
          "latency_ms_p90": 23333,
          "invalid_json_rate": 0.2,
          "by_field": {
            "supplier_name": 0.8,
            "supplier_id": 0.8,
            "vat_id": 0.8,
            "invoice_number": 0.7333,
            "payment_reference": 0.7333,
            "issue_date": 0.8,
            "tax_point_date": 0.8,
            "due_date": 0.8,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.8,
            "total": 0.8,
            "currency": 0.8,
            "bank_account": 0.8
          },
          "errors": {
            "invalid_json": 39,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "en",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9994,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 9.6483,
          "latency_ms_p50": 4074,
          "latency_ms_p90": 5519,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9933
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9982,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 8.8374,
          "latency_ms_p50": 3160,
          "latency_ms_p90": 3631,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9867,
            "vat_by_rate": 0.9933,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2,
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9982,
          "doc_accuracy": 0.98,
          "usd_per_1000": 3.8361,
          "latency_ms_p50": 3225,
          "latency_ms_p90": 4555,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.98
          },
          "errors": {
            "wrong_value": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.997,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 0.5074,
          "latency_ms_p50": 3435,
          "latency_ms_p90": 4678,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 0.9933,
            "payment_reference": 0.98,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9933
          },
          "errors": {
            "hallucinated": 3,
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9939,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.5188,
          "latency_ms_p50": 5737,
          "latency_ms_p90": 9427,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9933,
            "bank_account": 0.9733
          },
          "errors": {
            "wrong_value": 5,
            "hallucinated": 5
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9921,
          "doc_accuracy": 0.92,
          "usd_per_1000": 0.8663,
          "latency_ms_p50": 1700,
          "latency_ms_p90": 2034,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 0.9733,
            "due_date": 0.9933,
            "net_by_rate": 0.98,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 8,
            "hallucinated": 5
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9776,
          "doc_accuracy": 0.7933,
          "usd_per_1000": 4.3725,
          "latency_ms_p50": 2316,
          "latency_ms_p90": 5374,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9467,
            "issue_date": 0.9867,
            "due_date": 0.9933,
            "net_by_rate": 0.88,
            "vat_by_rate": 0.9867,
            "total": 1,
            "currency": 1,
            "bank_account": 0.96
          },
          "errors": {
            "wrong_value": 28,
            "hallucinated": 8,
            "wrong_rate": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9764,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 0.1675,
          "latency_ms_p50": 8869,
          "latency_ms_p90": 11901,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9933,
            "supplier_id": 0.98,
            "invoice_number": 1,
            "payment_reference": 0.9933,
            "issue_date": 0.9867,
            "due_date": 0.9867,
            "net_by_rate": 0.9533,
            "vat_by_rate": 0.9867,
            "total": 1,
            "currency": 1,
            "bank_account": 0.86
          },
          "errors": {
            "wrong_value": 38,
            "wrong_rate": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9558,
          "doc_accuracy": 0.58,
          "usd_per_1000": 0.4076,
          "latency_ms_p50": 1802,
          "latency_ms_p90": 2647,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9933,
            "invoice_number": 1,
            "payment_reference": 0.8733,
            "issue_date": 0.9733,
            "due_date": 0.9867,
            "net_by_rate": 0.72,
            "vat_by_rate": 0.9933,
            "total": 1,
            "currency": 0.9933,
            "bank_account": 0.98
          },
          "errors": {
            "wrong_rate": 39,
            "wrong_value": 20,
            "hallucinated": 13,
            "swapped_dates": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_rate",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8236,
          "doc_accuracy": 0.5733,
          "usd_per_1000": 1.1543,
          "latency_ms_p50": 17427,
          "latency_ms_p90": 37173,
          "invalid_json_rate": 0.0933,
          "by_field": {
            "supplier_name": 0.8533,
            "supplier_id": 0.8867,
            "invoice_number": 0.8533,
            "payment_reference": 0.7933,
            "issue_date": 0.8533,
            "due_date": 0.8533,
            "net_by_rate": 0.7,
            "vat_by_rate": 0.76,
            "total": 0.8533,
            "currency": 0.84,
            "bank_account": 0.8133
          },
          "errors": {
            "invalid_json": 154,
            "missing_field": 83,
            "wrong_rate": 23,
            "wrong_value": 18,
            "hallucinated": 13
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "en",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.5161,
          "latency_ms_p50": 7386,
          "latency_ms_p90": 11591,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 2.9193,
          "latency_ms_p50": 2292,
          "latency_ms_p90": 3558,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.1812,
          "latency_ms_p50": 7413,
          "latency_ms_p90": 10416,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.1805,
          "latency_ms_p50": 2651,
          "latency_ms_p90": 3772,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.2831,
          "latency_ms_p50": 3234,
          "latency_ms_p90": 5149,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.998,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 3.9142,
          "latency_ms_p50": 2368,
          "latency_ms_p90": 3335,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.998,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 2.2418,
          "latency_ms_p50": 1173,
          "latency_ms_p90": 1533,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9939,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.718,
          "latency_ms_p50": 962,
          "latency_ms_p90": 1043,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9556,
            "due_date": 0.9778,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 3
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9838,
          "doc_accuracy": 0.8444,
          "usd_per_1000": 0.1901,
          "latency_ms_p50": 1187,
          "latency_ms_p90": 1557,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.9778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 4,
            "hallucinated": 3,
            "wrong_rate": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9172,
          "doc_accuracy": 0.4889,
          "usd_per_1000": 0.9702,
          "latency_ms_p50": 18679,
          "latency_ms_p90": 42527,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.5111,
            "vat_by_rate": 0.6667,
            "total": 1,
            "currency": 0.9778,
            "bank_account": 1
          },
          "errors": {
            "wrong_rate": 19,
            "wrong_value": 9,
            "missing_field": 7,
            "hallucinated": 6
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_rate",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "en",
      "level": "clean",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 12.19,
          "latency_ms_p50": 3267,
          "latency_ms_p90": 3631,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 13.8084,
          "latency_ms_p50": 4193,
          "latency_ms_p90": 5907,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.996,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.9272,
          "latency_ms_p50": 1824,
          "latency_ms_p90": 2152,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9778,
            "issue_date": 0.9778,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.996,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 4.1338,
          "latency_ms_p50": 3533,
          "latency_ms_p90": 4775,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9556
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.996,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.7251,
          "latency_ms_p50": 3796,
          "latency_ms_p90": 5315,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9556,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9919,
          "doc_accuracy": 0.9111,
          "usd_per_1000": 0.4957,
          "latency_ms_p50": 4555,
          "latency_ms_p90": 8879,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "hallucinated": 3,
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9818,
          "doc_accuracy": 0.8444,
          "usd_per_1000": 0.1605,
          "latency_ms_p50": 9201,
          "latency_ms_p90": 12274,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.9778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 9
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9798,
          "doc_accuracy": 0.8222,
          "usd_per_1000": 5.248,
          "latency_ms_p50": 2282,
          "latency_ms_p90": 6512,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9778,
            "due_date": 1,
            "net_by_rate": 0.9333,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 6,
            "hallucinated": 3,
            "wrong_rate": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9556,
          "doc_accuracy": 0.5556,
          "usd_per_1000": 0.4985,
          "latency_ms_p50": 1736,
          "latency_ms_p90": 2116,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9778,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 0.9778,
            "due_date": 1,
            "net_by_rate": 0.6667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_rate": 15,
            "hallucinated": 4,
            "wrong_value": 3
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_rate",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.7717,
          "doc_accuracy": 0.6,
          "usd_per_1000": 1.1297,
          "latency_ms_p50": 15210,
          "latency_ms_p90": 28007,
          "invalid_json_rate": 0.1333,
          "by_field": {
            "supplier_name": 0.7778,
            "supplier_id": 0.8444,
            "invoice_number": 0.7778,
            "payment_reference": 0.7333,
            "issue_date": 0.7778,
            "due_date": 0.7778,
            "net_by_rate": 0.7778,
            "vat_by_rate": 0.7778,
            "total": 0.7778,
            "currency": 0.7778,
            "bank_account": 0.6889
          },
          "errors": {
            "invalid_json": 66,
            "missing_field": 38,
            "wrong_value": 5,
            "hallucinated": 2,
            "wrong_rate": 2
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "en",
      "level": "scan",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 10.6568,
          "latency_ms_p50": 3319,
          "latency_ms_p90": 3696,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 12.0443,
          "latency_ms_p50": 4313,
          "latency_ms_p90": 5469,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 45,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.998,
          "doc_accuracy": 0.9778,
          "usd_per_1000": 4.3571,
          "latency_ms_p50": 3241,
          "latency_ms_p90": 4550,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.996,
          "doc_accuracy": 0.9556,
          "usd_per_1000": 0.6262,
          "latency_ms_p50": 3740,
          "latency_ms_p90": 4870,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 0.9778,
            "payment_reference": 0.9778,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1,
            "hallucinated": 1
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9899,
          "doc_accuracy": 0.8889,
          "usd_per_1000": 0.5453,
          "latency_ms_p50": 5351,
          "latency_ms_p90": 8515,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9556,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9778,
            "bank_account": 0.9556
          },
          "errors": {
            "wrong_value": 3,
            "hallucinated": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9879,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.9301,
          "latency_ms_p50": 1791,
          "latency_ms_p90": 2032,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9111,
            "issue_date": 0.9778,
            "due_date": 1,
            "net_by_rate": 0.9778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 4,
            "wrong_value": 2
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9657,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 0.162,
          "latency_ms_p50": 9798,
          "latency_ms_p90": 11657,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9778,
            "supplier_id": 0.9556,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9556,
            "vat_by_rate": 0.9556,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7778
          },
          "errors": {
            "wrong_value": 17
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9596,
          "doc_accuracy": 0.6222,
          "usd_per_1000": 5.2725,
          "latency_ms_p50": 2837,
          "latency_ms_p90": 5374,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.9778,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 15,
            "hallucinated": 5
          },
          "n_cases": 45,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9354,
          "doc_accuracy": 0.3778,
          "usd_per_1000": 0.4992,
          "latency_ms_p50": 2335,
          "latency_ms_p90": 2788,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.8222,
            "issue_date": 0.9778,
            "due_date": 0.9778,
            "net_by_rate": 0.5556,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9778,
            "bank_account": 0.9778
          },
          "errors": {
            "wrong_rate": 18,
            "wrong_value": 7,
            "hallucinated": 6,
            "swapped_dates": 1
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "wrong_rate",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8141,
          "doc_accuracy": 0.6,
          "usd_per_1000": 1.3487,
          "latency_ms_p50": 17435,
          "latency_ms_p90": 36557,
          "invalid_json_rate": 0.0889,
          "by_field": {
            "supplier_name": 0.8222,
            "supplier_id": 0.8667,
            "invoice_number": 0.8222,
            "payment_reference": 0.7333,
            "issue_date": 0.8222,
            "due_date": 0.8222,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.8444,
            "total": 0.8222,
            "currency": 0.8,
            "bank_account": 0.8
          },
          "errors": {
            "invalid_json": 44,
            "missing_field": 38,
            "hallucinated": 5,
            "wrong_value": 3,
            "wrong_rate": 2
          },
          "n_cases": 45,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "en",
      "level": "photo",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.1305,
          "latency_ms_p50": 3362,
          "latency_ms_p90": 5045,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 15,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9939,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.5169,
          "latency_ms_p50": 5338,
          "latency_ms_p90": 7755,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9939,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 0.4784,
          "latency_ms_p50": 3709,
          "latency_ms_p90": 4204,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9939,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 9.0758,
          "latency_ms_p50": 4856,
          "latency_ms_p90": 5873,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9333
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9879,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 8.0913,
          "latency_ms_p50": 3176,
          "latency_ms_p90": 3589,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9879,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.9374,
          "latency_ms_p50": 1700,
          "latency_ms_p90": 2090,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9636,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 5.4385,
          "latency_ms_p50": 2815,
          "latency_ms_p90": 7530,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "due_date": 1,
            "net_by_rate": 0.6667,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 6
          },
          "n_cases": 15,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9333,
          "doc_accuracy": 0.4667,
          "usd_per_1000": 0.5128,
          "latency_ms_p50": 2597,
          "latency_ms_p90": 3148,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "due_date": 1,
            "net_by_rate": 0.6,
            "vat_by_rate": 0.9333,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 6,
            "wrong_rate": 5
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9212,
          "doc_accuracy": 0.4,
          "usd_per_1000": 0.1644,
          "latency_ms_p50": 10362,
          "latency_ms_p90": 11591,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.7333,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.6667
          },
          "errors": {
            "wrong_value": 12,
            "wrong_rate": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.7273,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 1.1969,
          "latency_ms_p50": 19570,
          "latency_ms_p90": 28157,
          "invalid_json_rate": 0.2667,
          "by_field": {
            "supplier_name": 0.7333,
            "supplier_id": 0.7333,
            "invoice_number": 0.7333,
            "payment_reference": 0.7333,
            "issue_date": 0.7333,
            "due_date": 0.7333,
            "net_by_rate": 0.7333,
            "vat_by_rate": 0.7333,
            "total": 0.7333,
            "currency": 0.7333,
            "bank_account": 0.6667
          },
          "errors": {
            "invalid_json": 44,
            "wrong_value": 1
          },
          "n_cases": 15,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "es",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9915,
          "doc_accuracy": 0.89,
          "usd_per_1000": 10.7592,
          "latency_ms_p50": 5692,
          "latency_ms_p90": 7507,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.98,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.92,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.99
          },
          "errors": {
            "missing_field": 8,
            "wrong_value": 3
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9908,
          "doc_accuracy": 0.88,
          "usd_per_1000": 0.5521,
          "latency_ms_p50": 3902,
          "latency_ms_p90": 5214,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.99,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.98,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.91
          },
          "errors": {
            "wrong_value": 10,
            "hallucinated": 2
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9825,
          "doc_accuracy": 0.7722,
          "usd_per_1000": 0.7056,
          "latency_ms_p50": 5820,
          "latency_ms_p90": 13530,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9367,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9747,
            "issue_date": 1,
            "tax_point_date": 0.9873,
            "due_date": 1,
            "net_by_rate": 0.8861,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9873
          },
          "errors": {
            "missing_field": 9,
            "hallucinated": 3,
            "lost_diacritics": 3,
            "wrong_value": 3
          },
          "n_cases": 79,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9823,
          "doc_accuracy": 0.79,
          "usd_per_1000": 1.0491,
          "latency_ms_p50": 1569,
          "latency_ms_p90": 2142,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.98,
            "supplier_id": 1,
            "vat_id": 0.83,
            "invoice_number": 1,
            "payment_reference": 0.99,
            "issue_date": 0.99,
            "tax_point_date": 0.98,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 19,
            "wrong_value": 4
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.97,
          "doc_accuracy": 0.67,
          "usd_per_1000": 4.7463,
          "latency_ms_p50": 2487,
          "latency_ms_p90": 3994,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.99,
            "supplier_id": 1,
            "vat_id": 0.94,
            "invoice_number": 1,
            "payment_reference": 0.93,
            "issue_date": 1,
            "tax_point_date": 0.93,
            "due_date": 0.97,
            "net_by_rate": 0.85,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 20,
            "missing_field": 18,
            "wrong_value": 1
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9592,
          "doc_accuracy": 0.95,
          "usd_per_1000": 7.8637,
          "latency_ms_p50": 5272,
          "latency_ms_p90": 9979,
          "invalid_json_rate": 0.04,
          "by_field": {
            "supplier_name": 0.95,
            "supplier_id": 0.96,
            "vat_id": 0.96,
            "invoice_number": 0.96,
            "payment_reference": 0.96,
            "issue_date": 0.96,
            "tax_point_date": 0.96,
            "due_date": 0.96,
            "net_by_rate": 0.96,
            "vat_by_rate": 0.96,
            "total": 0.96,
            "currency": 0.96,
            "bank_account": 0.96
          },
          "errors": {
            "invalid_json": 52,
            "wrong_value": 1
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.954,
          "doc_accuracy": 0.5402,
          "usd_per_1000": 0.2039,
          "latency_ms_p50": 8667,
          "latency_ms_p90": 11107,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.954,
            "supplier_id": 0.931,
            "vat_id": 0.9885,
            "invoice_number": 1,
            "payment_reference": 0.954,
            "issue_date": 0.9885,
            "tax_point_date": 1,
            "due_date": 0.977,
            "net_by_rate": 0.8161,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7931
          },
          "errors": {
            "wrong_value": 35,
            "missing_field": 16,
            "hallucinated": 1
          },
          "n_cases": 87,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9362,
          "doc_accuracy": 0.77,
          "usd_per_1000": 11.8874,
          "latency_ms_p50": 4045,
          "latency_ms_p90": 6706,
          "invalid_json_rate": 0.05,
          "by_field": {
            "supplier_name": 0.94,
            "supplier_id": 0.95,
            "vat_id": 0.95,
            "invoice_number": 0.95,
            "payment_reference": 0.94,
            "issue_date": 0.95,
            "tax_point_date": 0.95,
            "due_date": 0.95,
            "net_by_rate": 0.79,
            "vat_by_rate": 0.95,
            "total": 0.95,
            "currency": 0.95,
            "bank_account": 0.95
          },
          "errors": {
            "invalid_json": 65,
            "missing_field": 16,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 100,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9194,
          "doc_accuracy": 0.3095,
          "usd_per_1000": 0.4445,
          "latency_ms_p50": 2150,
          "latency_ms_p90": 2931,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9762,
            "supplier_id": 0.9881,
            "vat_id": 0.5357,
            "invoice_number": 1,
            "payment_reference": 0.881,
            "issue_date": 1,
            "tax_point_date": 0.6905,
            "due_date": 1,
            "net_by_rate": 0.9643,
            "vat_by_rate": 0.9762,
            "total": 0.9881,
            "currency": 1,
            "bank_account": 0.9524
          },
          "errors": {
            "hallucinated": 72,
            "wrong_value": 12,
            "swapped_dates": 3,
            "missing_field": 1
          },
          "n_cases": 84,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8792,
          "doc_accuracy": 0.74,
          "usd_per_1000": 1.3951,
          "latency_ms_p50": 22260,
          "latency_ms_p90": 38174,
          "invalid_json_rate": 0.1,
          "by_field": {
            "supplier_name": 0.86,
            "supplier_id": 0.89,
            "vat_id": 0.86,
            "invoice_number": 0.89,
            "payment_reference": 0.85,
            "issue_date": 0.89,
            "tax_point_date": 0.87,
            "due_date": 0.89,
            "net_by_rate": 0.86,
            "vat_by_rate": 0.9,
            "total": 0.89,
            "currency": 0.89,
            "bank_account": 0.89
          },
          "errors": {
            "invalid_json": 130,
            "missing_field": 14,
            "hallucinated": 10,
            "wrong_value": 2,
            "lost_diacritics": 1
          },
          "n_cases": 100,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "es",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 5.2156,
          "latency_ms_p50": 4740,
          "latency_ms_p90": 6344,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 0.2333,
          "latency_ms_p50": 2998,
          "latency_ms_p90": 4122,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 4.3284,
          "latency_ms_p50": 4496,
          "latency_ms_p90": 5754,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 1.1228,
          "latency_ms_p50": 26587,
          "latency_ms_p90": 46450,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9904,
          "doc_accuracy": 0.875,
          "usd_per_1000": 0.5999,
          "latency_ms_p50": 7319,
          "latency_ms_p90": 13076,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.875,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 24,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 3.0348,
          "latency_ms_p50": 1473,
          "latency_ms_p90": 1671,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8,
          "usd_per_1000": 0.9366,
          "latency_ms_p50": 1106,
          "latency_ms_p90": 1248,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 6
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9808,
          "doc_accuracy": 0.75,
          "usd_per_1000": 0.2358,
          "latency_ms_p50": 7074,
          "latency_ms_p90": 9658,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9643,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.7857,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 6,
            "hallucinated": 1
          },
          "n_cases": 28,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9615,
          "doc_accuracy": 0.5417,
          "usd_per_1000": 0.245,
          "latency_ms_p50": 1595,
          "latency_ms_p90": 1822,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8333,
            "invoice_number": 1,
            "payment_reference": 0.9583,
            "issue_date": 1,
            "tax_point_date": 0.75,
            "due_date": 1,
            "net_by_rate": 0.9583,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 11,
            "missing_field": 1
          },
          "n_cases": 24,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 6.8795,
          "latency_ms_p50": 4852,
          "latency_ms_p90": 7572,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.9,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.7667,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.9333
          },
          "errors": {
            "invalid_json": 26,
            "missing_field": 5,
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "es",
      "level": "clean",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.7926,
          "latency_ms_p50": 4324,
          "latency_ms_p90": 6150,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8667
          },
          "errors": {
            "wrong_value": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 15.4773,
          "latency_ms_p50": 6621,
          "latency_ms_p90": 7598,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9872,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 1.0878,
          "latency_ms_p50": 1833,
          "latency_ms_p90": 2152,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8333,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 5
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9829,
          "doc_accuracy": 0.7778,
          "usd_per_1000": 0.6743,
          "latency_ms_p50": 6196,
          "latency_ms_p90": 14735,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9259,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.8519,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 4,
            "wrong_value": 1,
            "lost_diacritics": 1
          },
          "n_cases": 27,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9692,
          "doc_accuracy": 0.6667,
          "usd_per_1000": 5.4235,
          "latency_ms_p50": 2491,
          "latency_ms_p90": 4179,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9667,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 0.9333,
            "due_date": 0.9667,
            "net_by_rate": 0.8333,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 6,
            "missing_field": 6
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9402,
          "doc_accuracy": 0.4074,
          "usd_per_1000": 0.1877,
          "latency_ms_p50": 8934,
          "latency_ms_p90": 11014,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.963,
            "supplier_id": 0.8519,
            "vat_id": 0.963,
            "invoice_number": 1,
            "payment_reference": 0.963,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.7778,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7037
          },
          "errors": {
            "wrong_value": 15,
            "missing_field": 6
          },
          "n_cases": 27,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9333,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 8.42,
          "latency_ms_p50": 6033,
          "latency_ms_p90": 10371,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.9333
          },
          "errors": {
            "invalid_json": 26
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9179,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 15.4133,
          "latency_ms_p50": 4202,
          "latency_ms_p90": 6706,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.7333,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.9333
          },
          "errors": {
            "invalid_json": 26,
            "missing_field": 6
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.8985,
          "doc_accuracy": 0.2,
          "usd_per_1000": 0.5201,
          "latency_ms_p50": 1958,
          "latency_ms_p90": 2590,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.44,
            "invoice_number": 1,
            "payment_reference": 0.84,
            "issue_date": 1,
            "tax_point_date": 0.6,
            "due_date": 1,
            "net_by_rate": 0.96,
            "vat_by_rate": 0.96,
            "total": 1,
            "currency": 1,
            "bank_account": 0.88
          },
          "errors": {
            "hallucinated": 28,
            "wrong_value": 5
          },
          "n_cases": 25,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8462,
          "doc_accuracy": 0.5667,
          "usd_per_1000": 1.4266,
          "latency_ms_p50": 19984,
          "latency_ms_p90": 38174,
          "invalid_json_rate": 0.1,
          "by_field": {
            "supplier_name": 0.8333,
            "supplier_id": 0.8667,
            "vat_id": 0.8,
            "invoice_number": 0.8667,
            "payment_reference": 0.8,
            "issue_date": 0.8667,
            "tax_point_date": 0.8,
            "due_date": 0.8667,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.9,
            "total": 0.8667,
            "currency": 0.8667,
            "bank_account": 0.8667
          },
          "errors": {
            "invalid_json": 39,
            "missing_field": 13,
            "hallucinated": 7,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "es",
      "level": "scan",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9872,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 0.6486,
          "latency_ms_p50": 3926,
          "latency_ms_p90": 5041,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9
          },
          "errors": {
            "wrong_value": 4,
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9872,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 12.714,
          "latency_ms_p50": 5833,
          "latency_ms_p90": 7288,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9667
          },
          "errors": {
            "missing_field": 3,
            "wrong_value": 2
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9744,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 1.0953,
          "latency_ms_p50": 1779,
          "latency_ms_p90": 2426,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 1,
            "vat_id": 0.8333,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9667,
            "tax_point_date": 0.9333,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 6,
            "wrong_value": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9731,
          "doc_accuracy": 0.65,
          "usd_per_1000": 0.635,
          "latency_ms_p50": 5363,
          "latency_ms_p90": 15461,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.95,
            "issue_date": 1,
            "tax_point_date": 0.95,
            "due_date": 1,
            "net_by_rate": 0.9,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.95
          },
          "errors": {
            "hallucinated": 2,
            "missing_field": 2,
            "wrong_value": 2,
            "lost_diacritics": 1
          },
          "n_cases": 20,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9538,
          "doc_accuracy": 0.5,
          "usd_per_1000": 5.4533,
          "latency_ms_p50": 2670,
          "latency_ms_p90": 5304,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 0.9,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 0.8667,
            "due_date": 0.9667,
            "net_by_rate": 0.8,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 10,
            "missing_field": 7,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9513,
          "doc_accuracy": 0.7667,
          "usd_per_1000": 13.5109,
          "latency_ms_p50": 3881,
          "latency_ms_p90": 6961,
          "invalid_json_rate": 0.0333,
          "by_field": {
            "supplier_name": 0.9333,
            "supplier_id": 0.9667,
            "vat_id": 0.9667,
            "invoice_number": 0.9667,
            "payment_reference": 0.9667,
            "issue_date": 0.9667,
            "tax_point_date": 0.9667,
            "due_date": 0.9667,
            "net_by_rate": 0.8,
            "vat_by_rate": 0.9667,
            "total": 0.9667,
            "currency": 0.9667,
            "bank_account": 0.9667
          },
          "errors": {
            "invalid_json": 13,
            "missing_field": 5,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9354,
          "doc_accuracy": 0.4,
          "usd_per_1000": 0.1882,
          "latency_ms_p50": 8870,
          "latency_ms_p90": 11584,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.92,
            "supplier_id": 0.92,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.92,
            "issue_date": 0.96,
            "tax_point_date": 1,
            "due_date": 0.96,
            "net_by_rate": 0.84,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.64
          },
          "errors": {
            "wrong_value": 17,
            "missing_field": 4
          },
          "n_cases": 25,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9308,
          "doc_accuracy": 0.9,
          "usd_per_1000": 10.2156,
          "latency_ms_p50": 5599,
          "latency_ms_p90": 13132,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.9,
            "supplier_id": 0.9333,
            "vat_id": 0.9333,
            "invoice_number": 0.9333,
            "payment_reference": 0.9333,
            "issue_date": 0.9333,
            "tax_point_date": 0.9333,
            "due_date": 0.9333,
            "net_by_rate": 0.9333,
            "vat_by_rate": 0.9333,
            "total": 0.9333,
            "currency": 0.9333,
            "bank_account": 0.9333
          },
          "errors": {
            "invalid_json": 26,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.8964,
          "doc_accuracy": 0.1538,
          "usd_per_1000": 0.5217,
          "latency_ms_p50": 2571,
          "latency_ms_p90": 3117,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9231,
            "supplier_id": 0.9615,
            "vat_id": 0.3462,
            "invoice_number": 1,
            "payment_reference": 0.8462,
            "issue_date": 1,
            "tax_point_date": 0.6923,
            "due_date": 1,
            "net_by_rate": 0.9615,
            "vat_by_rate": 0.9615,
            "total": 0.9615,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 28,
            "wrong_value": 6,
            "swapped_dates": 1
          },
          "n_cases": 26,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.7949,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 1.674,
          "latency_ms_p50": 20894,
          "latency_ms_p90": 35161,
          "invalid_json_rate": 0.2,
          "by_field": {
            "supplier_name": 0.7667,
            "supplier_id": 0.8,
            "vat_id": 0.8,
            "invoice_number": 0.8,
            "payment_reference": 0.8,
            "issue_date": 0.8,
            "tax_point_date": 0.8,
            "due_date": 0.8,
            "net_by_rate": 0.7667,
            "vat_by_rate": 0.8,
            "total": 0.8,
            "currency": 0.8,
            "bank_account": 0.8
          },
          "errors": {
            "invalid_json": 78,
            "missing_field": 1,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "es",
      "level": "photo",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 11.4632,
          "latency_ms_p50": 3863,
          "latency_ms_p90": 4692,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 10,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 7.0838,
          "latency_ms_p50": 7107,
          "latency_ms_p90": 12185,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 10,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9,
          "usd_per_1000": 10.0325,
          "latency_ms_p50": 6072,
          "latency_ms_p90": 8558,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8,
          "usd_per_1000": 1.1312,
          "latency_ms_p50": 1836,
          "latency_ms_p90": 2075,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8,
          "usd_per_1000": 0.4977,
          "latency_ms_p50": 4426,
          "latency_ms_p90": 5774,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9808,
          "doc_accuracy": 0.75,
          "usd_per_1000": 1.3048,
          "latency_ms_p50": 7650,
          "latency_ms_p90": 34555,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.875,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.875,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1,
            "lost_diacritics": 1
          },
          "n_cases": 8,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.967,
          "doc_accuracy": 0.7143,
          "usd_per_1000": 0.1953,
          "latency_ms_p50": 9250,
          "latency_ms_p90": 49805,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.8571,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 0.8571,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8571
          },
          "errors": {
            "wrong_value": 3
          },
          "n_cases": 7,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9615,
          "doc_accuracy": 0.6,
          "usd_per_1000": 5.7285,
          "latency_ms_p50": 2920,
          "latency_ms_p90": 4550,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 0.9,
            "due_date": 0.9,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 4,
            "missing_field": 1
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9316,
          "doc_accuracy": 0.4444,
          "usd_per_1000": 0.5438,
          "latency_ms_p50": 2839,
          "latency_ms_p90": 4022,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.5556,
            "invoice_number": 1,
            "payment_reference": 0.8889,
            "issue_date": 1,
            "tax_point_date": 0.7778,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8889
          },
          "errors": {
            "hallucinated": 5,
            "swapped_dates": 2,
            "wrong_value": 1
          },
          "n_cases": 9,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8769,
          "doc_accuracy": 0.6,
          "usd_per_1000": 1.2804,
          "latency_ms_p50": 17230,
          "latency_ms_p90": 47315,
          "invalid_json_rate": 0.1,
          "by_field": {
            "supplier_name": 0.8,
            "supplier_id": 0.9,
            "vat_id": 0.8,
            "invoice_number": 0.9,
            "payment_reference": 0.8,
            "issue_date": 0.9,
            "tax_point_date": 0.9,
            "due_date": 0.9,
            "net_by_rate": 0.9,
            "vat_by_rate": 0.9,
            "total": 0.9,
            "currency": 0.9,
            "bank_account": 0.9
          },
          "errors": {
            "invalid_json": 13,
            "hallucinated": 2,
            "lost_diacritics": 1
          },
          "n_cases": 10,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "fr",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9992,
          "doc_accuracy": 0.99,
          "usd_per_1000": 6.175,
          "latency_ms_p50": 5157,
          "latency_ms_p90": 9128,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.99,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9992,
          "doc_accuracy": 0.99,
          "usd_per_1000": 11.0514,
          "latency_ms_p50": 5874,
          "latency_ms_p90": 7599,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.99,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9938,
          "doc_accuracy": 0.9189,
          "usd_per_1000": 0.6481,
          "latency_ms_p50": 5737,
          "latency_ms_p90": 13604,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9865,
            "invoice_number": 1,
            "payment_reference": 0.9865,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9459
          },
          "errors": {
            "wrong_value": 5,
            "hallucinated": 1
          },
          "n_cases": 74,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.989,
          "doc_accuracy": 0.8571,
          "usd_per_1000": 0.5656,
          "latency_ms_p50": 3817,
          "latency_ms_p90": 5211,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9898,
            "supplier_id": 0.9898,
            "vat_id": 0.9796,
            "invoice_number": 0.9898,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9082
          },
          "errors": {
            "wrong_value": 14
          },
          "n_cases": 98,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9869,
          "doc_accuracy": 0.85,
          "usd_per_1000": 1.0656,
          "latency_ms_p50": 1514,
          "latency_ms_p90": 1844,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.98,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 0.99,
            "payment_reference": 0.91,
            "issue_date": 0.98,
            "tax_point_date": 0.98,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.99
          },
          "errors": {
            "hallucinated": 9,
            "wrong_value": 4,
            "swapped_dates": 4
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9865,
          "doc_accuracy": 0.9588,
          "usd_per_1000": 11.0212,
          "latency_ms_p50": 4366,
          "latency_ms_p90": 8310,
          "invalid_json_rate": 0.0103,
          "by_field": {
            "supplier_name": 0.9897,
            "supplier_id": 0.9897,
            "vat_id": 0.9897,
            "invoice_number": 0.9897,
            "payment_reference": 0.9794,
            "issue_date": 0.9897,
            "tax_point_date": 0.9897,
            "due_date": 0.9897,
            "net_by_rate": 0.9691,
            "vat_by_rate": 0.9897,
            "total": 0.9897,
            "currency": 0.9897,
            "bank_account": 0.9794
          },
          "errors": {
            "invalid_json": 13,
            "missing_field": 2,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 97,
          "passes_bar": true,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9815,
          "doc_accuracy": 0.81,
          "usd_per_1000": 4.8072,
          "latency_ms_p50": 2580,
          "latency_ms_p90": 3864,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.99,
            "supplier_id": 0.99,
            "vat_id": 0.98,
            "invoice_number": 1,
            "payment_reference": 0.94,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.95,
            "vat_by_rate": 0.98,
            "total": 0.99,
            "currency": 0.96,
            "bank_account": 0.98
          },
          "errors": {
            "wrong_value": 13,
            "hallucinated": 7,
            "missing_field": 4
          },
          "n_cases": 100,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9423,
          "doc_accuracy": 0.5,
          "usd_per_1000": 0.4465,
          "latency_ms_p50": 2103,
          "latency_ms_p90": 2719,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9605,
            "supplier_id": 0.9342,
            "vat_id": 0.9342,
            "invoice_number": 1,
            "payment_reference": 0.8553,
            "issue_date": 1,
            "tax_point_date": 0.9211,
            "due_date": 1,
            "net_by_rate": 0.9474,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9474,
            "bank_account": 0.75
          },
          "errors": {
            "wrong_value": 33,
            "hallucinated": 23,
            "swapped_dates": 1
          },
          "n_cases": 76,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9355,
          "doc_accuracy": 0.5057,
          "usd_per_1000": 0.2218,
          "latency_ms_p50": 7185,
          "latency_ms_p90": 10891,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9655,
            "supplier_id": 0.7701,
            "vat_id": 0.954,
            "invoice_number": 0.977,
            "payment_reference": 0.9655,
            "issue_date": 0.9655,
            "tax_point_date": 0.9425,
            "due_date": 0.9655,
            "net_by_rate": 0.931,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9885,
            "bank_account": 0.7356
          },
          "errors": {
            "wrong_value": 64,
            "missing_field": 7,
            "swapped_dates": 2
          },
          "n_cases": 87,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9062,
          "doc_accuracy": 0.82,
          "usd_per_1000": 1.2291,
          "latency_ms_p50": 19842,
          "latency_ms_p90": 35110,
          "invalid_json_rate": 0.05,
          "by_field": {
            "supplier_name": 0.91,
            "supplier_id": 0.91,
            "vat_id": 0.91,
            "invoice_number": 0.9,
            "payment_reference": 0.86,
            "issue_date": 0.91,
            "tax_point_date": 0.92,
            "due_date": 0.89,
            "net_by_rate": 0.91,
            "vat_by_rate": 0.94,
            "total": 0.91,
            "currency": 0.91,
            "bank_account": 0.9
          },
          "errors": {
            "invalid_json": 65,
            "missing_field": 49,
            "hallucinated": 5,
            "wrong_value": 3
          },
          "n_cases": 100,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "fr",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 6.227,
          "latency_ms_p50": 4870,
          "latency_ms_p90": 7771,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 29,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 5.0002,
          "latency_ms_p50": 3463,
          "latency_ms_p90": 6616,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.3417,
          "latency_ms_p50": 4249,
          "latency_ms_p90": 5222,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 3.1419,
          "latency_ms_p50": 1516,
          "latency_ms_p90": 1914,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 1.0013,
          "latency_ms_p50": 24352,
          "latency_ms_p90": 41640,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9973,
          "doc_accuracy": 0.9643,
          "usd_per_1000": 0.2335,
          "latency_ms_p50": 3081,
          "latency_ms_p90": 4731,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9643
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 28,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9963,
          "doc_accuracy": 0.9524,
          "usd_per_1000": 0.7155,
          "latency_ms_p50": 6122,
          "latency_ms_p90": 17391,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9524
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 21,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9945,
          "doc_accuracy": 0.9286,
          "usd_per_1000": 0.2581,
          "latency_ms_p50": 6489,
          "latency_ms_p90": 9401,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9643,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9643
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 28,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9,
          "usd_per_1000": 0.965,
          "latency_ms_p50": 1123,
          "latency_ms_p90": 1268,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 3
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9825,
          "doc_accuracy": 0.8182,
          "usd_per_1000": 0.2683,
          "latency_ms_p50": 1468,
          "latency_ms_p90": 1776,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9545,
            "invoice_number": 1,
            "payment_reference": 0.8636,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9545,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 5
          },
          "n_cases": 22,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "fr",
      "level": "clean",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 6.2689,
          "latency_ms_p50": 5347,
          "latency_ms_p90": 9205,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 15.4335,
          "latency_ms_p50": 6112,
          "latency_ms_p90": 7684,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9973,
          "doc_accuracy": 0.9655,
          "usd_per_1000": 13.2247,
          "latency_ms_p50": 4236,
          "latency_ms_p90": 8310,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9655,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 29,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.993,
          "doc_accuracy": 0.9091,
          "usd_per_1000": 0.5977,
          "latency_ms_p50": 4972,
          "latency_ms_p90": 9320,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9091
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 22,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8,
          "usd_per_1000": 1.0959,
          "latency_ms_p50": 1642,
          "latency_ms_p90": 2004,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 0.9667,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9667
          },
          "errors": {
            "wrong_value": 3,
            "hallucinated": 3
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9795,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 0.7835,
          "latency_ms_p50": 4250,
          "latency_ms_p90": 5459,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7667
          },
          "errors": {
            "wrong_value": 8
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9769,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 5.4665,
          "latency_ms_p50": 2604,
          "latency_ms_p90": 5998,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 0.9667,
            "vat_id": 0.9333,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9667,
            "bank_account": 0.9667
          },
          "errors": {
            "wrong_value": 5,
            "hallucinated": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9138,
          "doc_accuracy": 0.32,
          "usd_per_1000": 0.5145,
          "latency_ms_p50": 1999,
          "latency_ms_p90": 2562,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.96,
            "supplier_id": 0.88,
            "vat_id": 0.92,
            "invoice_number": 1,
            "payment_reference": 0.88,
            "issue_date": 1,
            "tax_point_date": 0.88,
            "due_date": 1,
            "net_by_rate": 0.88,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.92,
            "bank_account": 0.56
          },
          "errors": {
            "wrong_value": 17,
            "hallucinated": 11
          },
          "n_cases": 25,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9135,
          "doc_accuracy": 0.2917,
          "usd_per_1000": 0.2015,
          "latency_ms_p50": 7893,
          "latency_ms_p90": 10655,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9167,
            "supplier_id": 0.5833,
            "vat_id": 0.9583,
            "invoice_number": 0.9583,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 0.9167,
            "due_date": 1,
            "net_by_rate": 0.875,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.6667
          },
          "errors": {
            "wrong_value": 25,
            "missing_field": 2
          },
          "n_cases": 24,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8359,
          "doc_accuracy": 0.7667,
          "usd_per_1000": 1.2627,
          "latency_ms_p50": 17811,
          "latency_ms_p90": 31208,
          "invalid_json_rate": 0.1,
          "by_field": {
            "supplier_name": 0.8333,
            "supplier_id": 0.8333,
            "vat_id": 0.8333,
            "invoice_number": 0.8,
            "payment_reference": 0.8,
            "issue_date": 0.8333,
            "tax_point_date": 0.8667,
            "due_date": 0.8333,
            "net_by_rate": 0.8333,
            "vat_by_rate": 0.9,
            "total": 0.8333,
            "currency": 0.8333,
            "bank_account": 0.8333
          },
          "errors": {
            "invalid_json": 39,
            "missing_field": 22,
            "wrong_value": 2,
            "hallucinated": 1
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "fr",
      "level": "scan",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 13.6762,
          "latency_ms_p50": 6192,
          "latency_ms_p90": 7906,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 30,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9974,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 6.8685,
          "latency_ms_p50": 5924,
          "latency_ms_p90": 9433,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9936,
          "doc_accuracy": 0.9167,
          "usd_per_1000": 0.6538,
          "latency_ms_p50": 5804,
          "latency_ms_p90": 11754,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9583,
            "invoice_number": 1,
            "payment_reference": 0.9583,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 24,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9333,
          "usd_per_1000": 12.9166,
          "latency_ms_p50": 4476,
          "latency_ms_p90": 11382,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9667,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9667,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9667
          },
          "errors": {
            "missing_field": 1,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9897,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.6806,
          "latency_ms_p50": 4021,
          "latency_ms_p90": 5274,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.9333,
            "invoice_number": 0.9667,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9667
          },
          "errors": {
            "wrong_value": 4
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 1.1119,
          "latency_ms_p50": 1604,
          "latency_ms_p90": 1933,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9667,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 0.9667,
            "tax_point_date": 0.9667,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "hallucinated": 3,
            "swapped_dates": 2,
            "wrong_value": 1
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9744,
          "doc_accuracy": 0.7333,
          "usd_per_1000": 5.4967,
          "latency_ms_p50": 2960,
          "latency_ms_p90": 3864,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9,
            "bank_account": 0.9667
          },
          "errors": {
            "wrong_value": 4,
            "missing_field": 3,
            "hallucinated": 3
          },
          "n_cases": 30,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9336,
          "doc_accuracy": 0.4091,
          "usd_per_1000": 0.5178,
          "latency_ms_p50": 2526,
          "latency_ms_p90": 3038,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.9091,
            "supplier_id": 0.9091,
            "vat_id": 0.9545,
            "invoice_number": 1,
            "payment_reference": 0.8182,
            "issue_date": 1,
            "tax_point_date": 0.9091,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.9091,
            "bank_account": 0.7273
          },
          "errors": {
            "wrong_value": 13,
            "hallucinated": 6
          },
          "n_cases": 22,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9262,
          "doc_accuracy": 0.36,
          "usd_per_1000": 0.2052,
          "latency_ms_p50": 7696,
          "latency_ms_p90": 11658,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 0.96,
            "supplier_id": 0.72,
            "vat_id": 0.96,
            "invoice_number": 1,
            "payment_reference": 0.96,
            "issue_date": 0.96,
            "tax_point_date": 0.96,
            "due_date": 1,
            "net_by_rate": 0.92,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 0.96,
            "bank_account": 0.64
          },
          "errors": {
            "wrong_value": 19,
            "missing_field": 3,
            "swapped_dates": 2
          },
          "n_cases": 25,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.8615,
          "doc_accuracy": 0.7667,
          "usd_per_1000": 1.3844,
          "latency_ms_p50": 18975,
          "latency_ms_p90": 34797,
          "invalid_json_rate": 0.0667,
          "by_field": {
            "supplier_name": 0.8667,
            "supplier_id": 0.8667,
            "vat_id": 0.8667,
            "invoice_number": 0.8667,
            "payment_reference": 0.8,
            "issue_date": 0.8667,
            "tax_point_date": 0.8667,
            "due_date": 0.8333,
            "net_by_rate": 0.8667,
            "vat_by_rate": 0.9,
            "total": 0.8667,
            "currency": 0.8667,
            "bank_account": 0.8667
          },
          "errors": {
            "invalid_json": 26,
            "missing_field": 26,
            "hallucinated": 2
          },
          "n_cases": 30,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "invoices",
      "lang": "fr",
      "level": "photo",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 7.3376,
          "latency_ms_p50": 7206,
          "latency_ms_p90": 9278,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {},
          "n_cases": 10,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9,
          "usd_per_1000": 0.4973,
          "latency_ms_p50": 3715,
          "latency_ms_p90": 4945,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.9,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9923,
          "doc_accuracy": 0.9,
          "usd_per_1000": 10.16,
          "latency_ms_p50": 6752,
          "latency_ms_p90": 9928,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.989,
          "doc_accuracy": 0.8571,
          "usd_per_1000": 0.5851,
          "latency_ms_p50": 5820,
          "latency_ms_p90": 9139,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.8571
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 7,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9846,
          "doc_accuracy": 0.9,
          "usd_per_1000": 1.138,
          "latency_ms_p50": 1514,
          "latency_ms_p90": 1767,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 0.9,
            "tax_point_date": 0.9,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "swapped_dates": 2
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "swapped_dates",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9769,
          "doc_accuracy": 0.7,
          "usd_per_1000": 1.3457,
          "latency_ms_p50": 20764,
          "latency_ms_p90": 41787,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 0.9,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 0.9,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.9
          },
          "errors": {
            "wrong_value": 1,
            "hallucinated": 1,
            "missing_field": 1
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9692,
          "doc_accuracy": 0.8,
          "usd_per_1000": 5.757,
          "latency_ms_p50": 2932,
          "latency_ms_p90": 10840,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 1,
            "invoice_number": 1,
            "payment_reference": 1,
            "issue_date": 1,
            "tax_point_date": 1,
            "due_date": 1,
            "net_by_rate": 0.9,
            "vat_by_rate": 0.8,
            "total": 0.9,
            "currency": 1,
            "bank_account": 1
          },
          "errors": {
            "wrong_value": 4
          },
          "n_cases": 10,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9451,
          "doc_accuracy": 0.4286,
          "usd_per_1000": 0.54,
          "latency_ms_p50": 2653,
          "latency_ms_p90": 2957,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 1,
            "vat_id": 0.8571,
            "invoice_number": 1,
            "payment_reference": 0.8571,
            "issue_date": 1,
            "tax_point_date": 0.8571,
            "due_date": 1,
            "net_by_rate": 1,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.7143
          },
          "errors": {
            "wrong_value": 3,
            "hallucinated": 1,
            "swapped_dates": 1
          },
          "n_cases": 7,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.8889,
          "doc_accuracy": 0.8889,
          "usd_per_1000": 13.0511,
          "latency_ms_p50": 5444,
          "latency_ms_p90": 13861,
          "invalid_json_rate": 0.1111,
          "by_field": {
            "supplier_name": 0.8889,
            "supplier_id": 0.8889,
            "vat_id": 0.8889,
            "invoice_number": 0.8889,
            "payment_reference": 0.8889,
            "issue_date": 0.8889,
            "tax_point_date": 0.8889,
            "due_date": 0.8889,
            "net_by_rate": 0.8889,
            "vat_by_rate": 0.8889,
            "total": 0.8889,
            "currency": 0.8889,
            "bank_account": 0.8889
          },
          "errors": {
            "invalid_json": 13
          },
          "n_cases": 9,
          "passes_bar": false,
          "top_error": "invalid_json",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.8462,
          "doc_accuracy": 0.2,
          "usd_per_1000": 0.2105,
          "latency_ms_p50": 8460,
          "latency_ms_p90": 11644,
          "invalid_json_rate": 0,
          "by_field": {
            "supplier_name": 1,
            "supplier_id": 0.8,
            "vat_id": 0.8,
            "invoice_number": 0.9,
            "payment_reference": 0.8,
            "issue_date": 0.8,
            "tax_point_date": 0.8,
            "due_date": 0.7,
            "net_by_rate": 0.9,
            "vat_by_rate": 1,
            "total": 1,
            "currency": 1,
            "bank_account": 0.5
          },
          "errors": {
            "wrong_value": 18,
            "missing_field": 2
          },
          "n_cases": 10,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "cs",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9919,
          "doc_accuracy": 0.9267,
          "usd_per_1000": 3.0603,
          "latency_ms_p50": 2583,
          "latency_ms_p90": 3789,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.9333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 11
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9896,
          "doc_accuracy": 0.9067,
          "usd_per_1000": 3.1916,
          "latency_ms_p50": 2819,
          "latency_ms_p90": 4298,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.9267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 14
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9852,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.4453,
          "latency_ms_p50": 12968,
          "latency_ms_p90": 22193,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.8867,
            "customer_name": 0.9933,
            "order_number": 0.9933,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 20
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9815,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 4.9087,
          "latency_ms_p50": 4678,
          "latency_ms_p90": 7113,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8733,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 25
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9793,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 0.1781,
          "latency_ms_p50": 2554,
          "latency_ms_p90": 3123,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9867,
            "sentiment": 0.8267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 28
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9741,
          "doc_accuracy": 0.7733,
          "usd_per_1000": 0.2957,
          "latency_ms_p50": 4152,
          "latency_ms_p90": 7361,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9867,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 35
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9711,
          "doc_accuracy": 0.7467,
          "usd_per_1000": 0.1589,
          "latency_ms_p50": 2514,
          "latency_ms_p90": 5260,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8533,
            "sentiment": 0.8867,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 39
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9578,
          "doc_accuracy": 0.64,
          "usd_per_1000": 0.5091,
          "latency_ms_p50": 731,
          "latency_ms_p90": 823,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8,
            "sentiment": 0.84,
            "customer_name": 0.98,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 57
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9452,
          "doc_accuracy": 0.5533,
          "usd_per_1000": 1.9474,
          "latency_ms_p50": 822,
          "latency_ms_p90": 1112,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6333,
            "sentiment": 0.8733,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 74
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.877,
          "doc_accuracy": 0.3154,
          "usd_per_1000": 0.1177,
          "latency_ms_p50": 811,
          "latency_ms_p90": 1189,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5235,
            "sentiment": 0.7248,
            "customer_name": 0.9866,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.9799,
            "currency": 0.9799,
            "deadline": 0.698
          },
          "errors": {
            "wrong_value": 115,
            "hallucinated": 44,
            "missing_field": 6
          },
          "n_cases": 149,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "cs",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9919,
          "doc_accuracy": 0.9267,
          "usd_per_1000": 3.0603,
          "latency_ms_p50": 2583,
          "latency_ms_p90": 3789,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.9333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 11
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9896,
          "doc_accuracy": 0.9067,
          "usd_per_1000": 3.1916,
          "latency_ms_p50": 2819,
          "latency_ms_p90": 4298,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.9267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 14
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9852,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 0.4453,
          "latency_ms_p50": 12968,
          "latency_ms_p90": 22193,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.8867,
            "customer_name": 0.9933,
            "order_number": 0.9933,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 20
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9815,
          "doc_accuracy": 0.8333,
          "usd_per_1000": 4.9087,
          "latency_ms_p50": 4678,
          "latency_ms_p90": 7113,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8733,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 25
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9793,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 0.1781,
          "latency_ms_p50": 2554,
          "latency_ms_p90": 3123,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9867,
            "sentiment": 0.8267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 28
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9741,
          "doc_accuracy": 0.7733,
          "usd_per_1000": 0.2957,
          "latency_ms_p50": 4152,
          "latency_ms_p90": 7361,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9867,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 35
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9711,
          "doc_accuracy": 0.7467,
          "usd_per_1000": 0.1589,
          "latency_ms_p50": 2514,
          "latency_ms_p90": 5260,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8533,
            "sentiment": 0.8867,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 39
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9578,
          "doc_accuracy": 0.64,
          "usd_per_1000": 0.5091,
          "latency_ms_p50": 731,
          "latency_ms_p90": 823,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8,
            "sentiment": 0.84,
            "customer_name": 0.98,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 57
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9452,
          "doc_accuracy": 0.5533,
          "usd_per_1000": 1.9474,
          "latency_ms_p50": 822,
          "latency_ms_p90": 1112,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6333,
            "sentiment": 0.8733,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 74
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.877,
          "doc_accuracy": 0.3154,
          "usd_per_1000": 0.1177,
          "latency_ms_p50": 811,
          "latency_ms_p90": 1189,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5235,
            "sentiment": 0.7248,
            "customer_name": 0.9866,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.9799,
            "currency": 0.9799,
            "deadline": 0.698
          },
          "errors": {
            "wrong_value": 115,
            "hallucinated": 44,
            "missing_field": 6
          },
          "n_cases": 149,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "de",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9837,
          "doc_accuracy": 0.8533,
          "usd_per_1000": 2.743,
          "latency_ms_p50": 2862,
          "latency_ms_p90": 4287,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 1,
            "sentiment": 0.8533,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 22
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9822,
          "doc_accuracy": 0.84,
          "usd_per_1000": 2.729,
          "latency_ms_p50": 2939,
          "latency_ms_p90": 4115,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8667,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 24
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9785,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 5.1127,
          "latency_ms_p50": 4312,
          "latency_ms_p90": 10821,
          "invalid_json_rate": 0,
          "by_field": {
            "currency": 1,
            "deadline": 1,
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8467,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1
          },
          "errors": {
            "wrong_value": 29
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9785,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 0.3915,
          "latency_ms_p50": 12071,
          "latency_ms_p90": 18672,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8467,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 29
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9696,
          "doc_accuracy": 0.7267,
          "usd_per_1000": 0.1524,
          "latency_ms_p50": 2776,
          "latency_ms_p90": 3503,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.7333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 41
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9632,
          "doc_accuracy": 0.6759,
          "usd_per_1000": 0.3032,
          "latency_ms_p50": 4157,
          "latency_ms_p90": 7119,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9724,
            "sentiment": 0.6966,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 48
          },
          "n_cases": 145,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9511,
          "doc_accuracy": 0.62,
          "usd_per_1000": 0.1483,
          "latency_ms_p50": 3591,
          "latency_ms_p90": 5135,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.78,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 66
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9467,
          "doc_accuracy": 0.5667,
          "usd_per_1000": 0.4767,
          "latency_ms_p50": 766,
          "latency_ms_p90": 854,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.76,
            "sentiment": 0.7667,
            "customer_name": 0.9933,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 72
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9274,
          "doc_accuracy": 0.48,
          "usd_per_1000": 1.7705,
          "latency_ms_p50": 843,
          "latency_ms_p90": 1126,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5667,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 98
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.871,
          "doc_accuracy": 0.2937,
          "usd_per_1000": 0.1068,
          "latency_ms_p50": 809,
          "latency_ms_p90": 1105,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5524,
            "sentiment": 0.6503,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.965,
            "currency": 0.965,
            "deadline": 0.7063
          },
          "errors": {
            "wrong_value": 114,
            "hallucinated": 46,
            "missing_field": 6
          },
          "n_cases": 143,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "de",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9837,
          "doc_accuracy": 0.8533,
          "usd_per_1000": 2.743,
          "latency_ms_p50": 2862,
          "latency_ms_p90": 4287,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 1,
            "sentiment": 0.8533,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 22
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9822,
          "doc_accuracy": 0.84,
          "usd_per_1000": 2.729,
          "latency_ms_p50": 2939,
          "latency_ms_p90": 4115,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8667,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 24
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9785,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 5.1127,
          "latency_ms_p50": 4312,
          "latency_ms_p90": 10821,
          "invalid_json_rate": 0,
          "by_field": {
            "currency": 1,
            "deadline": 1,
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8467,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1
          },
          "errors": {
            "wrong_value": 29
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9785,
          "doc_accuracy": 0.8133,
          "usd_per_1000": 0.3915,
          "latency_ms_p50": 12071,
          "latency_ms_p90": 18672,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.96,
            "sentiment": 0.8467,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 29
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9696,
          "doc_accuracy": 0.7267,
          "usd_per_1000": 0.1524,
          "latency_ms_p50": 2776,
          "latency_ms_p90": 3503,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9933,
            "sentiment": 0.7333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 41
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9632,
          "doc_accuracy": 0.6759,
          "usd_per_1000": 0.3032,
          "latency_ms_p50": 4157,
          "latency_ms_p90": 7119,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9724,
            "sentiment": 0.6966,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 48
          },
          "n_cases": 145,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9511,
          "doc_accuracy": 0.62,
          "usd_per_1000": 0.1483,
          "latency_ms_p50": 3591,
          "latency_ms_p90": 5135,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.78,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 66
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9467,
          "doc_accuracy": 0.5667,
          "usd_per_1000": 0.4767,
          "latency_ms_p50": 766,
          "latency_ms_p90": 854,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.76,
            "sentiment": 0.7667,
            "customer_name": 0.9933,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 72
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9274,
          "doc_accuracy": 0.48,
          "usd_per_1000": 1.7705,
          "latency_ms_p50": 843,
          "latency_ms_p90": 1126,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5667,
            "sentiment": 0.78,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 98
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.871,
          "doc_accuracy": 0.2937,
          "usd_per_1000": 0.1068,
          "latency_ms_p50": 809,
          "latency_ms_p90": 1105,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.5524,
            "sentiment": 0.6503,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.965,
            "currency": 0.965,
            "deadline": 0.7063
          },
          "errors": {
            "wrong_value": 114,
            "hallucinated": 46,
            "missing_field": 6
          },
          "n_cases": 143,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "en",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9904,
          "doc_accuracy": 0.92,
          "usd_per_1000": 2.4361,
          "latency_ms_p50": 2793,
          "latency_ms_p90": 4032,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.9333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 13
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9874,
          "doc_accuracy": 0.8867,
          "usd_per_1000": 2.3857,
          "latency_ms_p50": 2161,
          "latency_ms_p90": 3312,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 1,
            "sentiment": 0.8867,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 17
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9822,
          "doc_accuracy": 0.84,
          "usd_per_1000": 4.0088,
          "latency_ms_p50": 3781,
          "latency_ms_p90": 5478,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8667,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 24
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9763,
          "doc_accuracy": 0.7867,
          "usd_per_1000": 0.2328,
          "latency_ms_p50": 3832,
          "latency_ms_p90": 6583,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9667,
            "sentiment": 0.82,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 32
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9748,
          "doc_accuracy": 0.7733,
          "usd_per_1000": 0.1347,
          "latency_ms_p50": 2529,
          "latency_ms_p90": 3081,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.7933,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 34
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9748,
          "doc_accuracy": 0.7867,
          "usd_per_1000": 0.3042,
          "latency_ms_p50": 8894,
          "latency_ms_p90": 15462,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8067,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 0.9933
          },
          "errors": {
            "wrong_value": 33,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9652,
          "doc_accuracy": 0.72,
          "usd_per_1000": 0.1207,
          "latency_ms_p50": 2210,
          "latency_ms_p90": 3749,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8533,
            "sentiment": 0.8333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 47
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.963,
          "doc_accuracy": 0.6867,
          "usd_per_1000": 0.4187,
          "latency_ms_p50": 715,
          "latency_ms_p90": 852,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.84,
            "sentiment": 0.8267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 50
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9489,
          "doc_accuracy": 0.5733,
          "usd_per_1000": 1.4742,
          "latency_ms_p50": 760,
          "latency_ms_p90": 1033,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6333,
            "sentiment": 0.9067,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 69
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.897,
          "doc_accuracy": 0.3733,
          "usd_per_1000": 0.0949,
          "latency_ms_p50": 807,
          "latency_ms_p90": 1020,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6133,
            "sentiment": 0.7133,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.9933,
            "currency": 0.9933,
            "deadline": 0.76
          },
          "errors": {
            "wrong_value": 103,
            "hallucinated": 34,
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "emails",
      "lang": "en",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9904,
          "doc_accuracy": 0.92,
          "usd_per_1000": 2.4361,
          "latency_ms_p50": 2793,
          "latency_ms_p90": 4032,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.9333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 13
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9874,
          "doc_accuracy": 0.8867,
          "usd_per_1000": 2.3857,
          "latency_ms_p50": 2161,
          "latency_ms_p90": 3312,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 1,
            "sentiment": 0.8867,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 17
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9822,
          "doc_accuracy": 0.84,
          "usd_per_1000": 4.0088,
          "latency_ms_p50": 3781,
          "latency_ms_p90": 5478,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8667,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 24
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9763,
          "doc_accuracy": 0.7867,
          "usd_per_1000": 0.2328,
          "latency_ms_p50": 3832,
          "latency_ms_p90": 6583,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9667,
            "sentiment": 0.82,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 32
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9748,
          "doc_accuracy": 0.7733,
          "usd_per_1000": 0.1347,
          "latency_ms_p50": 2529,
          "latency_ms_p90": 3081,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.98,
            "sentiment": 0.7933,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 34
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9748,
          "doc_accuracy": 0.7867,
          "usd_per_1000": 0.3042,
          "latency_ms_p50": 8894,
          "latency_ms_p90": 15462,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.9733,
            "sentiment": 0.8067,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 0.9933
          },
          "errors": {
            "wrong_value": 33,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9652,
          "doc_accuracy": 0.72,
          "usd_per_1000": 0.1207,
          "latency_ms_p50": 2210,
          "latency_ms_p90": 3749,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.8533,
            "sentiment": 0.8333,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 47
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.963,
          "doc_accuracy": 0.6867,
          "usd_per_1000": 0.4187,
          "latency_ms_p50": 715,
          "latency_ms_p90": 852,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.84,
            "sentiment": 0.8267,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 50
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9489,
          "doc_accuracy": 0.5733,
          "usd_per_1000": 1.4742,
          "latency_ms_p50": 760,
          "latency_ms_p90": 1033,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6333,
            "sentiment": 0.9067,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 1,
            "currency": 1,
            "deadline": 1
          },
          "errors": {
            "wrong_value": 69
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.897,
          "doc_accuracy": 0.3733,
          "usd_per_1000": 0.0949,
          "latency_ms_p50": 807,
          "latency_ms_p90": 1020,
          "invalid_json_rate": 0,
          "by_field": {
            "category": 1,
            "priority": 0.6133,
            "sentiment": 0.7133,
            "customer_name": 1,
            "order_number": 1,
            "invoice_number": 1,
            "amount": 0.9933,
            "currency": 0.9933,
            "deadline": 0.76
          },
          "errors": {
            "wrong_value": 103,
            "hallucinated": 34,
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "cs",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 6.0522,
          "latency_ms_p50": 3577,
          "latency_ms_p90": 4968,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.4195,
          "latency_ms_p50": 4273,
          "latency_ms_p90": 7378,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.0902,
          "latency_ms_p50": 2237,
          "latency_ms_p90": 3206,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.0563,
          "latency_ms_p50": 1183,
          "latency_ms_p90": 1537,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.27,
          "latency_ms_p50": 2557,
          "latency_ms_p90": 3083,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.6242,
          "latency_ms_p50": 2044,
          "latency_ms_p90": 2698,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.2541,
          "latency_ms_p50": 6019,
          "latency_ms_p90": 10114,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.6301,
          "latency_ms_p50": 13621,
          "latency_ms_p90": 20470,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 0.9867,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9985,
          "doc_accuracy": 0.98,
          "usd_per_1000": 0.8518,
          "latency_ms_p50": 929,
          "latency_ms_p90": 1073,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 1,
            "client_name": 0.9933,
            "client_id": 1,
            "contract_type": 0.9933,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2,
            "lost_diacritics": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9903,
          "doc_accuracy": 0.92,
          "usd_per_1000": 0.2268,
          "latency_ms_p50": 1154,
          "latency_ms_p90": 1460,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9533,
            "provider_id": 1,
            "client_name": 0.9533,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9667,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 16,
            "swapped_dates": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "cs",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 6.0522,
          "latency_ms_p50": 3577,
          "latency_ms_p90": 4968,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.4195,
          "latency_ms_p50": 4273,
          "latency_ms_p90": 7378,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.0902,
          "latency_ms_p50": 2237,
          "latency_ms_p90": 3206,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.0563,
          "latency_ms_p50": 1183,
          "latency_ms_p90": 1537,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.27,
          "latency_ms_p50": 2557,
          "latency_ms_p90": 3083,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.6242,
          "latency_ms_p50": 2044,
          "latency_ms_p90": 2698,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.2541,
          "latency_ms_p50": 6019,
          "latency_ms_p90": 10114,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.6301,
          "latency_ms_p50": 13621,
          "latency_ms_p90": 20470,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 0.9867,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9985,
          "doc_accuracy": 0.98,
          "usd_per_1000": 0.8518,
          "latency_ms_p50": 929,
          "latency_ms_p90": 1073,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 1,
            "client_name": 0.9933,
            "client_id": 1,
            "contract_type": 0.9933,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2,
            "lost_diacritics": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9903,
          "doc_accuracy": 0.92,
          "usd_per_1000": 0.2268,
          "latency_ms_p50": 1154,
          "latency_ms_p90": 1460,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9533,
            "provider_id": 1,
            "client_name": 0.9533,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9667,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 16,
            "swapped_dates": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "de",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.2663,
          "latency_ms_p50": 3021,
          "latency_ms_p90": 3846,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.3558,
          "latency_ms_p50": 2933,
          "latency_ms_p90": 4171,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 3.4159,
          "latency_ms_p50": 2618,
          "latency_ms_p90": 3849,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.9933,
            "client_name": 1,
            "client_id": 0.9933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9821,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 7.1056,
          "latency_ms_p50": 4322,
          "latency_ms_p90": 9847,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.8733,
            "client_name": 1,
            "client_id": 0.8933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 35
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.98,
          "doc_accuracy": 0.84,
          "usd_per_1000": 0.7911,
          "latency_ms_p50": 974,
          "latency_ms_p90": 1057,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.86,
            "client_name": 1,
            "client_id": 0.8867,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 0.9933,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 38,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.98,
          "doc_accuracy": 0.8733,
          "usd_per_1000": 0.7798,
          "latency_ms_p50": 19706,
          "latency_ms_p90": 38812,
          "invalid_json_rate": 0.0067,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 0.9467,
            "client_name": 0.9933,
            "client_id": 0.9467,
            "contract_type": 0.9933,
            "price": 0.9933,
            "currency": 0.9933,
            "price_period": 0.9933,
            "effective_date": 0.9933,
            "end_date": 0.9933,
            "notice_period_months": 0.9933,
            "penalty_pct_per_day": 0.9933,
            "governing_law": 0.9133
          },
          "errors": {
            "hallucinated": 14,
            "invalid_json": 13,
            "missing_field": 12
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9724,
          "doc_accuracy": 0.7724,
          "usd_per_1000": 0.5113,
          "latency_ms_p50": 4648,
          "latency_ms_p90": 9792,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.8138,
            "client_name": 1,
            "client_id": 0.8276,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 52
          },
          "n_cases": 145,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9713,
          "doc_accuracy": 0.7533,
          "usd_per_1000": 2.7624,
          "latency_ms_p50": 1249,
          "latency_ms_p90": 1606,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.82,
            "client_name": 1,
            "client_id": 0.8067,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 43,
            "missing_field": 9,
            "wrong_value": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9441,
          "doc_accuracy": 0.2733,
          "usd_per_1000": 0.2203,
          "latency_ms_p50": 2603,
          "latency_ms_p90": 6307,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 0.2733,
            "governing_law": 1
          },
          "errors": {
            "number_format": 109
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "number_format",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.94,
          "doc_accuracy": 0.4468,
          "usd_per_1000": 0.2057,
          "latency_ms_p50": 1226,
          "latency_ms_p90": 1614,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9858,
            "provider_id": 0.6241,
            "client_name": 0.9858,
            "client_id": 0.6312,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9929,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 99,
            "wrong_value": 11
          },
          "n_cases": 141,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "de",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.2663,
          "latency_ms_p50": 3021,
          "latency_ms_p90": 3846,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 4.3558,
          "latency_ms_p50": 2933,
          "latency_ms_p90": 4171,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 3.4159,
          "latency_ms_p50": 2618,
          "latency_ms_p90": 3849,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.9933,
            "client_name": 1,
            "client_id": 0.9933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9821,
          "doc_accuracy": 0.8667,
          "usd_per_1000": 7.1056,
          "latency_ms_p50": 4322,
          "latency_ms_p90": 9847,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.8733,
            "client_name": 1,
            "client_id": 0.8933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 35
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.98,
          "doc_accuracy": 0.84,
          "usd_per_1000": 0.7911,
          "latency_ms_p50": 974,
          "latency_ms_p90": 1057,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.86,
            "client_name": 1,
            "client_id": 0.8867,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 0.9933,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 38,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.98,
          "doc_accuracy": 0.8733,
          "usd_per_1000": 0.7798,
          "latency_ms_p50": 19706,
          "latency_ms_p90": 38812,
          "invalid_json_rate": 0.0067,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 0.9467,
            "client_name": 0.9933,
            "client_id": 0.9467,
            "contract_type": 0.9933,
            "price": 0.9933,
            "currency": 0.9933,
            "price_period": 0.9933,
            "effective_date": 0.9933,
            "end_date": 0.9933,
            "notice_period_months": 0.9933,
            "penalty_pct_per_day": 0.9933,
            "governing_law": 0.9133
          },
          "errors": {
            "hallucinated": 14,
            "invalid_json": 13,
            "missing_field": 12
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9724,
          "doc_accuracy": 0.7724,
          "usd_per_1000": 0.5113,
          "latency_ms_p50": 4648,
          "latency_ms_p90": 9792,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.8138,
            "client_name": 1,
            "client_id": 0.8276,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 52
          },
          "n_cases": 145,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9713,
          "doc_accuracy": 0.7533,
          "usd_per_1000": 2.7624,
          "latency_ms_p50": 1249,
          "latency_ms_p90": 1606,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 0.82,
            "client_name": 1,
            "client_id": 0.8067,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 43,
            "missing_field": 9,
            "wrong_value": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9441,
          "doc_accuracy": 0.2733,
          "usd_per_1000": 0.2203,
          "latency_ms_p50": 2603,
          "latency_ms_p90": 6307,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 0.2733,
            "governing_law": 1
          },
          "errors": {
            "number_format": 109
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "number_format",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.94,
          "doc_accuracy": 0.4468,
          "usd_per_1000": 0.2057,
          "latency_ms_p50": 1226,
          "latency_ms_p90": 1614,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9858,
            "provider_id": 0.6241,
            "client_name": 0.9858,
            "client_id": 0.6312,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9929,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "hallucinated": 99,
            "wrong_value": 11
          },
          "n_cases": 141,
          "passes_bar": false,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "en",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 5.039,
          "latency_ms_p50": 3395,
          "latency_ms_p90": 5890,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.3201,
          "latency_ms_p50": 3879,
          "latency_ms_p90": 6090,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.3295,
          "latency_ms_p50": 2483,
          "latency_ms_p90": 3443,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.692,
          "latency_ms_p50": 951,
          "latency_ms_p90": 1119,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 0.9933,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 2.4657,
          "latency_ms_p50": 2031,
          "latency_ms_p90": 3157,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 1,
            "client_name": 0.9933,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9979,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.201,
          "latency_ms_p50": 2715,
          "latency_ms_p90": 3285,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9867,
            "provider_id": 1,
            "client_name": 0.9867,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9964,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.18,
          "latency_ms_p50": 2507,
          "latency_ms_p90": 6551,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.98,
            "provider_id": 1,
            "client_name": 0.98,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 7
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.981,
          "doc_accuracy": 0.8733,
          "usd_per_1000": 2.3489,
          "latency_ms_p50": 1123,
          "latency_ms_p90": 1425,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.8733,
            "provider_id": 1,
            "client_name": 0.8867,
            "client_id": 0.9933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 36,
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9513,
          "doc_accuracy": 0.6933,
          "usd_per_1000": 0.6184,
          "latency_ms_p50": 14064,
          "latency_ms_p90": 23066,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.6933,
            "provider_id": 1,
            "client_name": 0.6933,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 0.9933,
            "governing_law": 0.9933
          },
          "errors": {
            "wrong_value": 92,
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9154,
          "doc_accuracy": 0.4333,
          "usd_per_1000": 0.17,
          "latency_ms_p50": 1127,
          "latency_ms_p90": 1407,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.4333,
            "provider_id": 1,
            "client_name": 0.4667,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 165
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "contracts",
      "lang": "en",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 5.039,
          "latency_ms_p50": 3395,
          "latency_ms_p90": 5890,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 0.3201,
          "latency_ms_p50": 3879,
          "latency_ms_p90": 6090,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 3.3295,
          "latency_ms_p50": 2483,
          "latency_ms_p90": 3443,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9995,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.692,
          "latency_ms_p50": 951,
          "latency_ms_p90": 1119,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 1,
            "provider_id": 1,
            "client_name": 1,
            "client_id": 1,
            "contract_type": 0.9933,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 2.4657,
          "latency_ms_p50": 2031,
          "latency_ms_p90": 3157,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9933,
            "provider_id": 1,
            "client_name": 0.9933,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9979,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.201,
          "latency_ms_p50": 2715,
          "latency_ms_p90": 3285,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.9867,
            "provider_id": 1,
            "client_name": 0.9867,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9964,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.18,
          "latency_ms_p50": 2507,
          "latency_ms_p90": 6551,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.98,
            "provider_id": 1,
            "client_name": 0.98,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 7
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.981,
          "doc_accuracy": 0.8733,
          "usd_per_1000": 2.3489,
          "latency_ms_p50": 1123,
          "latency_ms_p90": 1425,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.8733,
            "provider_id": 1,
            "client_name": 0.8867,
            "client_id": 0.9933,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 36,
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9513,
          "doc_accuracy": 0.6933,
          "usd_per_1000": 0.6184,
          "latency_ms_p50": 14064,
          "latency_ms_p90": 23066,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.6933,
            "provider_id": 1,
            "client_name": 0.6933,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 0.9933,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 0.9933,
            "governing_law": 0.9933
          },
          "errors": {
            "wrong_value": 92,
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9154,
          "doc_accuracy": 0.4333,
          "usd_per_1000": 0.17,
          "latency_ms_p50": 1127,
          "latency_ms_p90": 1407,
          "invalid_json_rate": 0,
          "by_field": {
            "provider_name": 0.4333,
            "provider_id": 1,
            "client_name": 0.4667,
            "client_id": 1,
            "contract_type": 1,
            "price": 1,
            "currency": 1,
            "price_period": 1,
            "effective_date": 1,
            "end_date": 1,
            "notice_period_months": 1,
            "penalty_pct_per_day": 1,
            "governing_law": 1
          },
          "errors": {
            "wrong_value": 165
          },
          "n_cases": 150,
          "passes_bar": false,
          "top_error": "wrong_value",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "cs",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.1189,
          "latency_ms_p50": 2907,
          "latency_ms_p90": 5834,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 1.3978,
          "latency_ms_p50": 833,
          "latency_ms_p90": 1063,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 0.3146,
          "latency_ms_p50": 4249,
          "latency_ms_p90": 8531,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.2218,
          "latency_ms_p50": 2531,
          "latency_ms_p90": 3300,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1195,
          "latency_ms_p50": 1846,
          "latency_ms_p90": 2421,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9952,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 2.1881,
          "latency_ms_p50": 2025,
          "latency_ms_p90": 2981,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3,
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9886,
          "doc_accuracy": 0.92,
          "usd_per_1000": 4.0557,
          "latency_ms_p50": 3178,
          "latency_ms_p90": 4965,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9667,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 0.98,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 8,
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9875,
          "doc_accuracy": 0.9195,
          "usd_per_1000": 0.0926,
          "latency_ms_p50": 889,
          "latency_ms_p90": 1251,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9396,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9866,
            "birth_dates": 0.9866,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 8,
            "missing_field": 5
          },
          "n_cases": 149,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.981,
          "doc_accuracy": 0.88,
          "usd_per_1000": 0.4542,
          "latency_ms_p50": 793,
          "latency_ms_p90": 926,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9133,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 0.96,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 14,
            "missing_field": 5,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9695,
          "doc_accuracy": 0.8867,
          "usd_per_1000": 0.4062,
          "latency_ms_p50": 9622,
          "latency_ms_p90": 17243,
          "invalid_json_rate": 0.0067,
          "by_field": {
            "person_names": 0.9067,
            "email_addresses": 0.98,
            "phone_numbers": 0.9733,
            "postal_addresses": 0.9667,
            "birth_dates": 0.98,
            "national_ids": 0.9933,
            "bank_accounts": 0.9867
          },
          "errors": {
            "wrong_value": 13,
            "invalid_json": 7,
            "hallucinated": 6,
            "missing_field": 6
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "cs",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.1189,
          "latency_ms_p50": 2907,
          "latency_ms_p90": 5834,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 1.3978,
          "latency_ms_p50": 833,
          "latency_ms_p90": 1063,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 0.3146,
          "latency_ms_p50": 4249,
          "latency_ms_p90": 8531,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.2218,
          "latency_ms_p50": 2531,
          "latency_ms_p90": 3300,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1195,
          "latency_ms_p50": 1846,
          "latency_ms_p90": 2421,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9952,
          "doc_accuracy": 0.9667,
          "usd_per_1000": 2.1881,
          "latency_ms_p50": 2025,
          "latency_ms_p90": 2981,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3,
            "hallucinated": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9886,
          "doc_accuracy": 0.92,
          "usd_per_1000": 4.0557,
          "latency_ms_p50": 3178,
          "latency_ms_p90": 4965,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9667,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 0.98,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 8,
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9875,
          "doc_accuracy": 0.9195,
          "usd_per_1000": 0.0926,
          "latency_ms_p50": 889,
          "latency_ms_p90": 1251,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9396,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9866,
            "birth_dates": 0.9866,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 8,
            "missing_field": 5
          },
          "n_cases": 149,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.981,
          "doc_accuracy": 0.88,
          "usd_per_1000": 0.4542,
          "latency_ms_p50": 793,
          "latency_ms_p90": 926,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9133,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 0.96,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 14,
            "missing_field": 5,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9695,
          "doc_accuracy": 0.8867,
          "usd_per_1000": 0.4062,
          "latency_ms_p50": 9622,
          "latency_ms_p90": 17243,
          "invalid_json_rate": 0.0067,
          "by_field": {
            "person_names": 0.9067,
            "email_addresses": 0.98,
            "phone_numbers": 0.9733,
            "postal_addresses": 0.9667,
            "birth_dates": 0.98,
            "national_ids": 0.9933,
            "bank_accounts": 0.9867
          },
          "errors": {
            "wrong_value": 13,
            "invalid_json": 7,
            "hallucinated": 6,
            "missing_field": 6
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "wrong_value",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "de",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.3835,
          "latency_ms_p50": 916,
          "latency_ms_p90": 1233,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.1019,
          "latency_ms_p50": 1876,
          "latency_ms_p90": 5172,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9865,
          "usd_per_1000": 0.3055,
          "latency_ms_p50": 3669,
          "latency_ms_p90": 6950,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9865,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2
          },
          "n_cases": 148,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 4.404,
          "latency_ms_p50": 3660,
          "latency_ms_p90": 6811,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 0.9933,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.1196,
          "latency_ms_p50": 2343,
          "latency_ms_p90": 3078,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.1268,
          "latency_ms_p50": 2597,
          "latency_ms_p90": 3732,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.4543,
          "latency_ms_p50": 813,
          "latency_ms_p90": 988,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9733,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1172,
          "latency_ms_p50": 2247,
          "latency_ms_p90": 3094,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.3589,
          "latency_ms_p50": 12189,
          "latency_ms_p90": 18612,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 0.9933,
            "phone_numbers": 0.9933,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9908,
          "doc_accuracy": 0.9353,
          "usd_per_1000": 0.0899,
          "latency_ms_p50": 943,
          "latency_ms_p90": 1272,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9496,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9928,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 0.9928
          },
          "errors": {
            "hallucinated": 7,
            "missing_field": 1,
            "wrong_value": 1
          },
          "n_cases": 139,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "de",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 1,
          "doc_accuracy": 1,
          "usd_per_1000": 1.3835,
          "latency_ms_p50": 916,
          "latency_ms_p90": 1233,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {},
          "n_cases": 150,
          "passes_bar": true,
          "top_error": null,
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.1019,
          "latency_ms_p50": 1876,
          "latency_ms_p90": 5172,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9981,
          "doc_accuracy": 0.9865,
          "usd_per_1000": 0.3055,
          "latency_ms_p50": 3669,
          "latency_ms_p90": 6950,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9865,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2
          },
          "n_cases": 148,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 4.404,
          "latency_ms_p50": 3660,
          "latency_ms_p90": 6811,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 0.9933,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.1196,
          "latency_ms_p50": 2343,
          "latency_ms_p90": 3078,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.1268,
          "latency_ms_p50": 2597,
          "latency_ms_p90": 3732,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.4543,
          "latency_ms_p50": 813,
          "latency_ms_p90": 988,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9733,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 1,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1172,
          "latency_ms_p50": 2247,
          "latency_ms_p90": 3094,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9733,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.3589,
          "latency_ms_p50": 12189,
          "latency_ms_p90": 18612,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 0.9933,
            "phone_numbers": 0.9933,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9908,
          "doc_accuracy": 0.9353,
          "usd_per_1000": 0.0899,
          "latency_ms_p50": 943,
          "latency_ms_p90": 1272,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9496,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9928,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 0.9928
          },
          "errors": {
            "hallucinated": 7,
            "missing_field": 1,
            "wrong_value": 1
          },
          "n_cases": 139,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "en",
      "level": "all",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.0845,
          "latency_ms_p50": 2396,
          "latency_ms_p90": 5354,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 3.5004,
          "latency_ms_p50": 2455,
          "latency_ms_p90": 4158,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 1.8639,
          "latency_ms_p50": 1944,
          "latency_ms_p90": 2646,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 1.147,
          "latency_ms_p50": 799,
          "latency_ms_p90": 1039,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.0916,
          "latency_ms_p50": 2598,
          "latency_ms_p90": 4610,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.2854,
          "latency_ms_p50": 3776,
          "latency_ms_p90": 6845,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1058,
          "latency_ms_p50": 1797,
          "latency_ms_p90": 2673,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 2,
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9952,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.4074,
          "latency_ms_p50": 776,
          "latency_ms_p90": 969,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 0.9933,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 0.9867,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9942,
          "doc_accuracy": 0.9597,
          "usd_per_1000": 0.0772,
          "latency_ms_p50": 834,
          "latency_ms_p90": 1225,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9732,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9866,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3,
            "missing_field": 3
          },
          "n_cases": 149,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9924,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.2791,
          "latency_ms_p50": 7481,
          "latency_ms_p90": 12254,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 0.9933,
            "phone_numbers": 0.9933,
            "postal_addresses": 0.9867,
            "birth_dates": 0.9933,
            "national_ids": 0.9933,
            "bank_accounts": 0.9933
          },
          "errors": {
            "hallucinated": 5,
            "wrong_value": 2,
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    },
    {
      "task": "pii",
      "lang": "en",
      "level": "text",
      "bar": 0.95,
      "rows": [
        {
          "model_id": "google/gemma-4-31b-it",
          "field_accuracy": 0.999,
          "doc_accuracy": 0.9933,
          "usd_per_1000": 0.0845,
          "latency_ms_p50": 2396,
          "latency_ms_p90": 5354,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9933,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemma 4 31B",
          "vendor": "Google"
        },
        {
          "model_id": "anthropic/claude-sonnet-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 3.5004,
          "latency_ms_p50": 2455,
          "latency_ms_p90": 4158,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Claude Sonnet 5",
          "vendor": "Anthropic"
        },
        {
          "model_id": "google/gemini-3.8-flash",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 1.8639,
          "latency_ms_p50": 1944,
          "latency_ms_p90": 2646,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.8 Flash",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-medium-3-5",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 1.147,
          "latency_ms_p50": 799,
          "latency_ms_p90": 1039,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Mistral Medium 3.5",
          "vendor": "Mistral"
        },
        {
          "model_id": "openai/gpt-6-sol",
          "field_accuracy": 0.9971,
          "doc_accuracy": 0.98,
          "usd_per_1000": 2.0916,
          "latency_ms_p50": 2598,
          "latency_ms_p90": 4610,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.98,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 3
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "GPT-6 Sol",
          "vendor": "OpenAI"
        },
        {
          "model_id": "deepseek/deepseek-v4.1-flash",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.2854,
          "latency_ms_p50": 3776,
          "latency_ms_p90": 6845,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 2,
            "hallucinated": 1,
            "wrong_value": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "DeepSeek V4.1 Flash",
          "vendor": "DeepSeek"
        },
        {
          "model_id": "openai/gpt-6-luna",
          "field_accuracy": 0.9962,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.1058,
          "latency_ms_p50": 1797,
          "latency_ms_p90": 2673,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9867,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 2,
            "missing_field": 2
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "GPT-6 Luna",
          "vendor": "OpenAI"
        },
        {
          "model_id": "google/gemini-3.5-flash-lite",
          "field_accuracy": 0.9952,
          "doc_accuracy": 0.9733,
          "usd_per_1000": 0.4074,
          "latency_ms_p50": 776,
          "latency_ms_p90": 969,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 1,
            "email_addresses": 0.9933,
            "phone_numbers": 1,
            "postal_addresses": 0.9867,
            "birth_dates": 0.9867,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "missing_field": 4,
            "hallucinated": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "missing_field",
          "model_name": "Gemini 3.5 Flash Lite",
          "vendor": "Google"
        },
        {
          "model_id": "mistralai/mistral-small-2603",
          "field_accuracy": 0.9942,
          "doc_accuracy": 0.9597,
          "usd_per_1000": 0.0772,
          "latency_ms_p50": 834,
          "latency_ms_p90": 1225,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9732,
            "email_addresses": 1,
            "phone_numbers": 1,
            "postal_addresses": 0.9866,
            "birth_dates": 1,
            "national_ids": 1,
            "bank_accounts": 1
          },
          "errors": {
            "hallucinated": 3,
            "missing_field": 3
          },
          "n_cases": 149,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Mistral Small 4",
          "vendor": "Mistral"
        },
        {
          "model_id": "qwen/qwen3.8-flash",
          "field_accuracy": 0.9924,
          "doc_accuracy": 0.9867,
          "usd_per_1000": 0.2791,
          "latency_ms_p50": 7481,
          "latency_ms_p90": 12254,
          "invalid_json_rate": 0,
          "by_field": {
            "person_names": 0.9933,
            "email_addresses": 0.9933,
            "phone_numbers": 0.9933,
            "postal_addresses": 0.9867,
            "birth_dates": 0.9933,
            "national_ids": 0.9933,
            "bank_accounts": 0.9933
          },
          "errors": {
            "hallucinated": 5,
            "wrong_value": 2,
            "missing_field": 1
          },
          "n_cases": 150,
          "passes_bar": true,
          "top_error": "hallucinated",
          "model_name": "Qwen3.8 Flash",
          "vendor": "Qwen"
        }
      ]
    }
  ]
}