Leaderboard
Which model does your task best
Sorted by the share of correct fields. Cost is the real token spend scaled to 1,000 documents. Bar: 95% correct fields; below it, manual fixes cost more than a cheaper model saves.
Every model gets the same prompt, with no per-model tuning. What that means →
Compare two models side by side →
Invoice extraction · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5 AnthropicEU data | 100.00% | $9.41 | 0 | 3.6 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 99.95% | $4.26 | 7 | 3.5 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 99.95% | $10.79 | 7 | 4.4 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.90% | $0.50 | 13 | 5.0 sec | |
| ||||||
| 5 | Claude Haiku 5.5New AnthropicEU data | 99.90% | $0.62 | 13 | 3.4 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 99.90% | $10.12 | 13 | 3.9 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.69% | $1.02 | 33 | 4.5 sec | |
| ||||||
| 8 | Gemini 3.5 Flash Lite GoogleEU data | 99.49% | $1.03 | 67 | 1.8 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 99.28% | $0.51 | 67 | 3.2 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 97.08% | $5.69 | 113 | 26.3 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 96.10% | $0.21 | 333 | 10.4 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 93.64% | $0.44 | 440 | 2.0 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 93.49% | $4.69 | 367 | 2.6 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 91.79% | $1.02 | 233 | 17.0 sec | |
| ||||||
Invoice extraction · Czech · PDF text (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.21 | 0 | 2.6 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.37 | 0 | 2.3 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.50 | 0 | 6.0 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.59 | 0 | 4.8 sec | |
| ||||||
| 5 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.91 | 0 | 1.1 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $3.46 | 0 | 2.8 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $4.02 | 0 | 2.8 sec | |
| ||||||
| 8 | Mistral Large 4New MistralEU data | 100.00% | $4.42 | 0 | 21.2 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 100.00% | $4.87 | 0 | 2.7 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 99.83% | $2.97 | 22 | 1.5 sec | |
| ||||||
| 11 | Claude Sonnet 5.5New AnthropicEU data | 99.83% | $6.32 | 22 | 4.5 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 99.49% | $0.24 | 67 | 9.0 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 96.75% | $0.26 | 244 | 1.4 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 91.97% | $0.74 | 467 | 19.9 sec | |
| ||||||
Invoice extraction · Czech · Clean image (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.49 | 0 | 4.4 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 100.00% | $4.61 | 0 | 4.0 sec | |
| ||||||
| 3 | Claude Sonnet 5 AnthropicEU data | 100.00% | $13.20 | 0 | 3.8 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $14.28 | 0 | 4.3 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $14.57 | 0 | 3.9 sec | |
| ||||||
| 6 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.82 | 22 | 3.9 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.83% | $1.39 | 22 | 4.6 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 99.66% | $0.73 | 44 | 3.6 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 99.49% | $1.07 | 67 | 2.0 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 96.24% | $5.85 | 156 | 22.9 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 95.38% | $0.20 | 378 | 11.0 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 92.65% | $1.11 | 156 | 15.9 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 92.31% | $0.51 | 556 | 1.9 sec | |
| ||||||
| 14 | Mistral Medium 3.5 MistralEU data | 92.14% | $5.37 | 400 | 2.5 sec | |
| ||||||
Invoice extraction · Czech · Scan (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5 AnthropicEU data | 100.00% | $10.49 | 0 | 3.7 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $11.99 | 0 | 4.4 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.68 | 22 | 3.6 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 99.83% | $4.61 | 22 | 3.5 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 99.83% | $11.89 | 22 | 4.2 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.66% | $0.49 | 44 | 4.8 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.32% | $1.14 | 67 | 4.3 sec | |
| ||||||
| 8 | Gemini 3.5 Flash Lite GoogleEU data | 98.97% | $1.08 | 133 | 1.8 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 98.29% | $0.59 | 133 | 3.4 sec | |
| ||||||
| 10 | Gemma 4 31B Google · open weights | 94.87% | $0.20 | 444 | 11.9 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 94.53% | $6.67 | 156 | 30.5 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 92.14% | $0.51 | 533 | 2.6 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 90.94% | $1.17 | 111 | 15.4 sec | |
| ||||||
| 14 | Mistral Medium 3.5 MistralEU data | 89.06% | $5.41 | 644 | 2.9 sec | |
| ||||||
Invoice extraction · Czech · Photo (15 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.54 | 0 | 5.4 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.60 | 0 | 3.4 sec | |
| ||||||
| 3 | Gemini 3.8 Flash | 100.00% | $4.53 | 0 | 3.4 sec | |
| ||||||
| 4 | Claude Sonnet 5 AnthropicEU data | 100.00% | $8.45 | 0 | 3.9 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $10.14 | 0 | 4.3 sec | |
| ||||||
| 6 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $0.90 | 67 | 4.5 sec | |
| ||||||
| 7 | Gemini 3.5 Flash Lite GoogleEU data | 99.49% | $1.11 | 67 | 1.8 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 99.49% | $9.76 | 67 | 4.9 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 98.97% | $0.47 | 133 | 3.3 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 98.46% | $6.03 | 200 | 30.9 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 92.82% | $0.54 | 400 | 2.9 sec | |
| ||||||
| 12 | Mistral Medium 3.5 MistralEU data | 91.79% | $5.65 | 467 | 3.0 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 91.79% | $0.20 | 667 | 11.0 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 91.28% | $1.11 | 133 | 15.4 sec | |
| ||||||
Invoice extraction · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $1.03 | 0 | 1.5 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 99.95% | $10.65 | 7 | 4.9 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 99.95% | $11.21 | 7 | 6.0 sec | |
| ||||||
| 4 | Claude Haiku 5.5New AnthropicEU data | 99.85% | $0.71 | 20 | 3.9 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 99.54% | $1.01 | 53 | 4.4 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.54% | $0.57 | 60 | 6.4 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 99.23% | $0.54 | 87 | 3.7 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 98.97% | $10.42 | 133 | 3.8 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 97.49% | $6.72 | 87 | 35.3 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 97.38% | $4.72 | 253 | 2.5 sec | |
| ||||||
| 11 | Gemini 3.8 Flash | 97.23% | $7.15 | 40 | 4.7 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 94.87% | $0.20 | 360 | 7.7 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 94.82% | $0.44 | 487 | 2.0 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 93.49% | $1.12 | 127 | 19.0 sec | |
| ||||||
Invoice extraction · German · PDF text (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.22 | 0 | 3.0 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.45 | 0 | 2.8 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.49 | 0 | 6.7 sec | |
| ||||||
| 4 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.91 | 0 | 1.1 sec | |
| ||||||
| 5 | Qwen3.8 Flash Qwen | 100.00% | $0.99 | 0 | 27.8 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 100.00% | $3.85 | 0 | 3.1 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $6.38 | 0 | 3.9 sec | |
| ||||||
| 8 | GLM-5.3 FlashNew Z.ai · open weights | 99.83% | $0.52 | 22 | 4.3 sec | |
| ||||||
| 9 | Gemma 4 31B Google · open weights | 99.49% | $0.23 | 67 | 7.0 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 99.49% | $2.96 | 67 | 1.4 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 99.32% | $5.89 | 89 | 3.3 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 97.95% | $0.25 | 200 | 1.3 sec | |
| ||||||
| 13 | Gemini 3.8 Flash | 97.78% | $6.24 | 22 | 3.9 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 97.44% | $5.89 | 67 | 33.1 sec | |
| ||||||
Invoice extraction · German · Clean image (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $1.07 | 0 | 1.7 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $15.26 | 0 | 5.6 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.87 | 22 | 4.3 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 99.83% | $13.98 | 22 | 6.0 sec | |
| ||||||
| 5 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.66% | $0.55 | 44 | 5.7 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.66% | $0.77 | 44 | 4.1 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.15% | $1.33 | 89 | 4.5 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 98.97% | $13.27 | 133 | 3.8 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 97.09% | $6.87 | 111 | 34.3 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 96.58% | $5.43 | 311 | 2.5 sec | |
| ||||||
| 11 | Gemini 3.8 Flash | 95.21% | $8.81 | 89 | 5.1 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 93.85% | $0.19 | 489 | 7.4 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 93.68% | $0.51 | 622 | 1.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 88.38% | $1.26 | 178 | 16.7 sec | |
| ||||||
Invoice extraction · German · Scan (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $1.08 | 0 | 1.6 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $13.49 | 0 | 6.4 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.81 | 22 | 4.2 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.83% | $13.15 | 22 | 5.1 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $1.19 | 67 | 4.4 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.97% | $0.61 | 133 | 7.2 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 98.63% | $11.88 | 178 | 3.8 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 98.29% | $0.66 | 178 | 4.1 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 97.78% | $6.65 | 22 | 5.0 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 97.09% | $7.10 | 111 | 40.6 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 96.92% | $1.15 | 133 | 16.6 sec | |
| ||||||
| 12 | Mistral Medium 3.5 MistralEU data | 96.92% | $5.45 | 333 | 2.8 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 92.99% | $0.52 | 644 | 2.4 sec | |
| ||||||
| 14 | Gemma 4 31B Google · open weights | 91.79% | $0.19 | 489 | 8.7 sec | |
| ||||||
Invoice extraction · German · Photo (15 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.93 | 0 | 4.4 sec | |
| ||||||
| 2 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $1.11 | 0 | 1.6 sec | |
| ||||||
| 3 | Gemini 3.8 Flash | 100.00% | $6.45 | 0 | 4.8 sec | |
| ||||||
| 4 | Mistral Large 4New MistralEU data | 100.00% | $7.59 | 0 | 37.7 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $9.73 | 0 | 5.4 sec | |
| ||||||
| 6 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $10.55 | 0 | 5.8 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 99.49% | $0.68 | 67 | 4.2 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.49% | $0.70 | 67 | 6.4 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 98.97% | $11.05 | 133 | 3.8 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 98.46% | $0.50 | 200 | 3.5 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 94.87% | $5.68 | 400 | 2.7 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 94.36% | $0.54 | 467 | 2.5 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 93.33% | $0.19 | 467 | 9.1 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 78.97% | $0.99 | 333 | 15.6 sec | |
| ||||||
Invoice extraction · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.61 | 0 | 3.4 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $9.54 | 0 | 3.5 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 99.94% | $0.95 | 7 | 4.2 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.94% | $9.65 | 7 | 4.1 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.82% | $8.84 | 13 | 3.2 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.82% | $3.84 | 20 | 3.2 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 99.70% | $0.51 | 33 | 3.4 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.39% | $0.52 | 67 | 5.7 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 99.21% | $0.87 | 80 | 1.7 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 97.76% | $4.37 | 207 | 2.3 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 97.64% | $0.17 | 187 | 8.9 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 95.58% | $0.41 | 420 | 1.8 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 95.39% | $6.65 | 227 | 29.0 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 82.36% | $1.15 | 427 | 17.4 sec | |
| ||||||
Invoice extraction · English (US) · PDF text (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.18 | 0 | 2.7 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.18 | 0 | 7.4 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.35 | 0 | 2.4 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.52 | 0 | 7.4 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.53 | 0 | 4.8 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $2.92 | 0 | 2.3 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $3.28 | 0 | 3.2 sec | |
| ||||||
| 8 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $4.86 | 0 | 2.4 sec | |
| ||||||
| 9 | Mistral Medium 3.5 MistralEU data | 99.80% | $2.24 | 22 | 1.2 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 99.80% | $3.91 | 22 | 2.4 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 99.39% | $0.72 | 44 | 1.0 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 98.38% | $0.19 | 156 | 1.2 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 97.58% | $5.25 | 44 | 30.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 91.72% | $0.97 | 511 | 18.7 sec | |
| ||||||
Invoice extraction · English (US) · Clean image (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.75 | 0 | 3.6 sec | |
| ||||||
| 2 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.23 | 0 | 4.0 sec | |
| ||||||
| 3 | Claude Sonnet 5 AnthropicEU data | 100.00% | $12.19 | 0 | 3.3 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $12.58 | 0 | 3.6 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $13.81 | 0 | 4.2 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.60% | $0.73 | 44 | 3.8 sec | |
| ||||||
| 7 | Gemini 3.5 Flash Lite GoogleEU data | 99.60% | $0.93 | 44 | 1.8 sec | |
| ||||||
| 8 | Gemini 3.8 Flash | 99.60% | $4.13 | 44 | 3.5 sec | |
| ||||||
| 9 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.19% | $0.50 | 89 | 4.6 sec | |
| ||||||
| 10 | Gemma 4 31B Google · open weights | 98.18% | $0.16 | 156 | 9.2 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 97.98% | $5.25 | 178 | 2.3 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 95.56% | $0.50 | 444 | 1.7 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 95.35% | $6.57 | 244 | 24.2 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 77.17% | $1.13 | 400 | 15.2 sec | |
| ||||||
Invoice extraction · English (US) · Scan (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.71 | 0 | 3.9 sec | |
| ||||||
| 2 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.08 | 0 | 4.0 sec | |
| ||||||
| 3 | Claude Sonnet 5 AnthropicEU data | 100.00% | $10.66 | 0 | 3.3 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $11.43 | 0 | 3.5 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $12.04 | 0 | 4.3 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.80% | $4.36 | 22 | 3.2 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 99.60% | $0.63 | 44 | 3.7 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.99% | $0.55 | 111 | 5.4 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 98.79% | $0.93 | 133 | 1.8 sec | |
| ||||||
| 10 | Gemma 4 31B Google · open weights | 96.57% | $0.16 | 267 | 9.8 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 95.96% | $5.27 | 378 | 2.8 sec | |
| ||||||
| 12 | Mistral Large 4New MistralEU data | 95.15% | $7.92 | 311 | 31.4 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 93.54% | $0.50 | 622 | 2.3 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 81.41% | $1.35 | 400 | 17.4 sec | |
| ||||||
Invoice extraction · English (US) · Photo (15 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.64 | 0 | 3.9 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 100.00% | $4.13 | 0 | 3.4 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $8.83 | 0 | 3.4 sec | |
| ||||||
| 4 | GPT-6 Luna OpenAIEU data | 99.39% | $0.48 | 67 | 3.7 sec | |
| ||||||
| 5 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.39% | $0.52 | 67 | 5.3 sec | |
| ||||||
| 6 | GLM-5.3 FlashNew Z.ai · open weights | 99.39% | $0.90 | 67 | 4.1 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.39% | $9.08 | 67 | 4.9 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 98.79% | $8.09 | 67 | 3.2 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 98.79% | $0.94 | 133 | 1.7 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 96.36% | $5.44 | 333 | 2.8 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 93.33% | $0.51 | 533 | 2.6 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 92.12% | $0.16 | 600 | 10.4 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 89.70% | $7.24 | 467 | 31.5 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 72.73% | $1.20 | 333 | 19.6 sec | |
| ||||||
Invoice extraction · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $10.99 | 0 | 5.7 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 99.44% | $0.73 | 73 | 4.4 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 99.38% | $1.06 | 73 | 5.0 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.23% | $10.79 | 100 | 5.9 sec | |
| ||||||
| 5 | GPT-6 Luna OpenAIEU data | 99.13% | $0.55 | 113 | 3.9 sec | |
| ||||||
| 6 | Claude Sonnet 5 AnthropicEU data | 98.46% | $11.51 | 187 | 4.2 sec | |
| ||||||
| 7 | Gemini 3.5 Flash Lite GoogleEU data | 98.21% | $1.05 | 220 | 1.6 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.21% | $0.65 | 227 | 6.3 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 97.28% | $7.84 | 33 | 5.7 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 96.87% | $4.75 | 333 | 2.5 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 96.51% | $7.49 | 120 | 41.3 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 94.62% | $0.20 | 493 | 7.4 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 92.36% | $0.44 | 673 | 2.1 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 88.87% | $1.38 | 253 | 22.0 sec | |
| ||||||
Invoice extraction · Spanish · PDF text (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 100.00% | $5.74 | 0 | 4.7 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $6.28 | 0 | 4.7 sec | |
| ||||||
| 3 | GPT-6 Luna OpenAIEU data | 99.83% | $0.23 | 22 | 3.0 sec | |
| ||||||
| 4 | Qwen3.8 Flash Qwen | 99.83% | $1.03 | 22 | 25.6 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 99.66% | $0.65 | 44 | 4.9 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 99.66% | $4.26 | 44 | 4.5 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 99.49% | $0.48 | 67 | 3.2 sec | |
| ||||||
| 8 | Mistral Large 4New MistralEU data | 99.49% | $5.90 | 67 | 37.9 sec | |
| ||||||
| 9 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.97% | $0.55 | 133 | 7.1 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 98.97% | $3.01 | 133 | 1.4 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 98.80% | $0.93 | 156 | 1.1 sec | |
| ||||||
| 12 | Claude Sonnet 5 AnthropicEU data | 98.80% | $6.50 | 156 | 4.9 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 98.46% | $0.23 | 200 | 6.6 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 95.90% | $0.25 | 467 | 1.6 sec | |
| ||||||
Invoice extraction · Spanish · Clean image (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $13.97 | 0 | 5.3 sec | |
| ||||||
| 2 | GLM-5.3 FlashNew Z.ai · open weights | 99.15% | $1.31 | 89 | 5.0 sec | |
| ||||||
| 3 | GPT-6 Luna OpenAIEU data | 99.15% | $0.79 | 111 | 4.3 sec | |
| ||||||
| 4 | Claude Haiku 5.5New AnthropicEU data | 98.97% | $0.90 | 133 | 4.7 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 98.97% | $15.49 | 133 | 6.7 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 98.29% | $1.09 | 222 | 1.8 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 98.29% | $15.12 | 222 | 4.4 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.78% | $0.63 | 289 | 6.3 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 96.92% | $7.79 | 133 | 39.3 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 96.92% | $5.44 | 356 | 2.5 sec | |
| ||||||
| 11 | Gemini 3.8 Flash | 95.56% | $8.64 | 44 | 6.0 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 93.85% | $0.19 | 578 | 7.9 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 90.77% | $0.52 | 756 | 2.0 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 87.18% | $1.50 | 356 | 20.0 sec | |
| ||||||
Invoice extraction · Spanish · Scan (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $12.89 | 0 | 6.0 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 99.66% | $0.82 | 44 | 4.9 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $1.29 | 67 | 5.1 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.15% | $12.82 | 111 | 6.1 sec | |
| ||||||
| 5 | GPT-6 Luna OpenAIEU data | 98.63% | $0.65 | 178 | 4.1 sec | |
| ||||||
| 6 | Claude Sonnet 5 AnthropicEU data | 98.46% | $13.18 | 178 | 3.9 sec | |
| ||||||
| 7 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.78% | $0.64 | 267 | 5.7 sec | |
| ||||||
| 8 | Gemini 3.5 Flash Lite GoogleEU data | 97.44% | $1.10 | 289 | 1.7 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 95.38% | $9.38 | 67 | 6.4 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 95.38% | $5.47 | 467 | 2.7 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 92.48% | $8.96 | 156 | 46.6 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 91.97% | $0.19 | 667 | 8.6 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 90.26% | $0.52 | 822 | 2.6 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 83.25% | $1.58 | 311 | 20.8 sec | |
| ||||||
Invoice extraction · Spanish · Photo (15 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.72 | 0 | 4.9 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 100.00% | $7.07 | 0 | 6.9 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $10.43 | 0 | 6.4 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 98.97% | $0.85 | 133 | 4.7 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 98.97% | $10.15 | 133 | 6.2 sec | |
| ||||||
| 6 | Mistral Large 4New MistralEU data | 98.46% | $6.93 | 133 | 33.8 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 98.46% | $0.50 | 200 | 4.4 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.46% | $1.04 | 200 | 7.3 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 98.46% | $1.13 | 200 | 1.7 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 97.95% | $10.71 | 200 | 4.0 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 94.87% | $5.71 | 467 | 3.0 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 93.33% | $0.19 | 600 | 7.4 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 92.82% | $0.54 | 600 | 2.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 77.95% | $1.51 | 467 | 18.3 sec | |
| ||||||
Invoice extraction · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 99.95% | $6.07 | 7 | 5.3 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 99.95% | $10.71 | 7 | 4.9 sec | |
| ||||||
| 3 | GPT-6 Sol OpenAIEU data | 99.90% | $11.00 | 13 | 6.0 sec | |
| ||||||
| 4 | Claude Haiku 5.5New AnthropicEU data | 99.85% | $0.77 | 20 | 4.6 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.69% | $10.69 | 33 | 4.4 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.54% | $0.60 | 60 | 6.2 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.08% | $0.98 | 107 | 4.7 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 99.08% | $0.56 | 120 | 3.9 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 98.97% | $1.07 | 120 | 1.5 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 98.56% | $4.82 | 153 | 2.6 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 94.87% | $0.44 | 473 | 2.1 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 93.54% | $0.22 | 493 | 6.5 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 90.46% | $1.28 | 167 | 19.4 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 88.97% | $8.71 | 347 | 43.4 sec | |
| ||||||
Invoice extraction · French · PDF text (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.52 | 0 | 3.4 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $4.28 | 0 | 4.2 sec | |
| ||||||
| 3 | Gemini 3.8 Flash | 100.00% | $4.95 | 0 | 3.9 sec | |
| ||||||
| 4 | Claude Sonnet 5 AnthropicEU data | 100.00% | $6.16 | 0 | 4.6 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $6.26 | 0 | 3.0 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.83% | $0.23 | 22 | 3.1 sec | |
| ||||||
| 7 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.83% | $0.59 | 22 | 6.1 sec | |
| ||||||
| 8 | Qwen3.8 Flash Qwen | 99.83% | $1.06 | 22 | 24.4 sec | |
| ||||||
| 9 | Mistral Medium 3.5 MistralEU data | 99.83% | $3.13 | 22 | 1.5 sec | |
| ||||||
| 10 | Gemma 4 31B Google · open weights | 99.66% | $0.26 | 44 | 5.9 sec | |
| ||||||
| 11 | GLM-5.3 FlashNew Z.ai · open weights | 99.66% | $0.58 | 44 | 5.0 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.49% | $0.96 | 67 | 1.1 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 98.63% | $0.26 | 156 | 1.4 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 97.44% | $6.23 | 67 | 40.3 sec | |
| ||||||
Invoice extraction · French · Clean image (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 100.00% | $6.33 | 0 | 5.5 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $15.44 | 0 | 6.6 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.93 | 22 | 4.9 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 99.83% | $13.55 | 22 | 5.1 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.66% | $13.17 | 44 | 4.3 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.49% | $0.58 | 67 | 6.1 sec | |
| ||||||
| 7 | Gemini 3.5 Flash Lite GoogleEU data | 98.63% | $1.10 | 178 | 1.7 sec | |
| ||||||
| 8 | Mistral Medium 3.5 MistralEU data | 98.46% | $5.49 | 178 | 2.6 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 98.46% | $0.79 | 200 | 4.3 sec | |
| ||||||
| 10 | GLM-5.3 FlashNew Z.ai · open weights | 97.95% | $1.27 | 222 | 4.4 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 93.33% | $0.52 | 578 | 2.0 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 91.28% | $0.20 | 667 | 6.9 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 87.86% | $9.95 | 467 | 42.6 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 83.08% | $1.29 | 222 | 17.7 sec | |
| ||||||
Invoice extraction · French · Scan (45 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $12.55 | 0 | 5.0 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $13.54 | 0 | 6.2 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.83% | $0.87 | 22 | 4.9 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 99.83% | $6.56 | 22 | 5.7 sec | |
| ||||||
| 5 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.66% | $0.62 | 44 | 6.1 sec | |
| ||||||
| 6 | Claude Sonnet 5 AnthropicEU data | 99.49% | $12.49 | 44 | 4.4 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $1.13 | 67 | 4.4 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 98.97% | $0.68 | 133 | 4.0 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 98.80% | $1.12 | 133 | 1.6 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 97.61% | $5.52 | 267 | 2.9 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 93.16% | $0.52 | 622 | 2.6 sec | |
| ||||||
| 12 | Gemma 4 31B Google · open weights | 92.14% | $0.21 | 667 | 6.5 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 88.21% | $1.48 | 200 | 18.3 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 82.91% | $9.63 | 511 | 47.6 sec | |
| ||||||
Invoice extraction · French · Photo (15 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 100.00% | $7.13 | 0 | 6.4 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $9.97 | 0 | 4.6 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 99.49% | $0.76 | 67 | 4.9 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $0.89 | 67 | 4.0 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.49% | $11.43 | 67 | 5.4 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 98.97% | $1.14 | 67 | 1.6 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 98.97% | $0.50 | 133 | 3.8 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 98.97% | $10.23 | 133 | 6.9 sec | |
| ||||||
| 9 | DeepSeek V4.1 Flash DeepSeek · open weights | 98.46% | $0.61 | 200 | 7.7 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 97.95% | $5.76 | 133 | 2.9 sec | |
| ||||||
| 11 | Mistral Small 4 Mistral · open weightsEU data | 93.33% | $0.54 | 667 | 2.7 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 91.28% | $1.32 | 333 | 19.8 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 86.15% | $0.21 | 800 | 7.2 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 85.13% | $9.74 | 333 | 49.7 sec | |
| ||||||
Email triage · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 99.19% | $3.06 | 73 | 2.6 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 98.96% | $3.19 | 93 | 2.8 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 98.89% | $0.52 | 100 | 4.7 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 98.67% | $3.92 | 120 | 2.0 sec | |
| ||||||
| 5 | Qwen3.8 Flash Qwen | 98.52% | $0.45 | 133 | 13.0 sec | |
| ||||||
| 6 | Mistral Large 4New MistralEU data | 98.15% | $4.41 | 107 | 28.4 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 98.15% | $4.91 | 167 | 4.7 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 97.93% | $0.18 | 187 | 2.6 sec | |
| ||||||
| 9 | Claude Haiku 5.5New AnthropicEU data | 97.63% | $0.29 | 200 | 1.9 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.41% | $0.30 | 227 | 4.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 97.11% | $0.16 | 253 | 2.5 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 95.78% | $0.51 | 360 | 0.7 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.52% | $1.95 | 447 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 87.63% | $0.12 | 687 | 0.8 sec | |
| ||||||
Email triage · Czech · Email (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 99.19% | $3.06 | 73 | 2.6 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 98.96% | $3.19 | 93 | 2.8 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 98.89% | $0.52 | 100 | 4.7 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 98.67% | $3.92 | 120 | 2.0 sec | |
| ||||||
| 5 | Qwen3.8 Flash Qwen | 98.52% | $0.45 | 133 | 13.0 sec | |
| ||||||
| 6 | Mistral Large 4New MistralEU data | 98.15% | $4.41 | 107 | 28.4 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 98.15% | $4.91 | 167 | 4.7 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 97.93% | $0.18 | 187 | 2.6 sec | |
| ||||||
| 9 | Claude Haiku 5.5New AnthropicEU data | 97.63% | $0.29 | 200 | 1.9 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.41% | $0.30 | 227 | 4.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 97.11% | $0.16 | 253 | 2.5 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 95.78% | $0.51 | 360 | 0.7 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.52% | $1.95 | 447 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 87.63% | $0.12 | 687 | 0.8 sec | |
| ||||||
Email triage · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 98.37% | $2.74 | 147 | 2.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 98.22% | $2.73 | 160 | 2.9 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 98.00% | $4.21 | 167 | 1.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 98.00% | $0.43 | 180 | 3.8 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 97.85% | $3.78 | 180 | 21.7 sec | |
| ||||||
| 6 | Qwen3.8 Flash Qwen | 97.85% | $0.39 | 187 | 12.1 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 97.85% | $5.11 | 187 | 4.3 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 97.11% | $0.32 | 253 | 2.1 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 96.96% | $0.15 | 273 | 2.8 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.37% | $0.30 | 320 | 4.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 95.11% | $0.15 | 380 | 3.6 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 94.67% | $0.48 | 433 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 92.74% | $1.77 | 520 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 87.26% | $0.11 | 707 | 0.8 sec | |
| ||||||
Email triage · German · Email (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | 98.37% | $2.74 | 147 | 2.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 98.22% | $2.73 | 160 | 2.9 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 98.00% | $4.21 | 167 | 1.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 98.00% | $0.43 | 180 | 3.8 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 97.85% | $3.78 | 180 | 21.7 sec | |
| ||||||
| 6 | Qwen3.8 Flash Qwen | 97.85% | $0.39 | 187 | 12.1 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 97.85% | $5.11 | 187 | 4.3 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 97.11% | $0.32 | 253 | 2.1 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 96.96% | $0.15 | 273 | 2.8 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.37% | $0.30 | 320 | 4.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 95.11% | $0.15 | 380 | 3.6 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 94.67% | $0.48 | 433 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 92.74% | $1.77 | 520 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 87.26% | $0.11 | 707 | 0.8 sec | |
| ||||||
Email triage · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 99.41% | $0.33 | 53 | 3.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 99.04% | $2.44 | 80 | 2.8 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 98.89% | $3.21 | 100 | 2.0 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 98.74% | $2.39 | 113 | 2.2 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 98.74% | $3.52 | 113 | 19.8 sec | |
| ||||||
| 6 | Claude Sonnet 5 AnthropicEU data | 98.22% | $4.01 | 160 | 3.8 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 98.15% | $0.23 | 160 | 1.7 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.63% | $0.23 | 213 | 3.8 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 97.48% | $0.30 | 213 | 8.9 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 97.48% | $0.13 | 227 | 2.5 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 96.52% | $0.12 | 280 | 2.2 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 96.30% | $0.42 | 313 | 0.7 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.89% | $1.47 | 427 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 89.70% | $0.09 | 627 | 0.8 sec | |
| ||||||
Email triage · English (US) · Email (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 99.41% | $0.33 | 53 | 3.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 99.04% | $2.44 | 80 | 2.8 sec | |
| ||||||
| 3 | Claude Sonnet 5.5New AnthropicEU data | 98.89% | $3.21 | 100 | 2.0 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 98.74% | $2.39 | 113 | 2.2 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 98.74% | $3.52 | 113 | 19.8 sec | |
| ||||||
| 6 | Claude Sonnet 5 AnthropicEU data | 98.22% | $4.01 | 160 | 3.8 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 98.15% | $0.23 | 160 | 1.7 sec | |
| ||||||
| 8 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.63% | $0.23 | 213 | 3.8 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 97.48% | $0.30 | 213 | 8.9 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 97.48% | $0.13 | 227 | 2.5 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 96.52% | $0.12 | 280 | 2.2 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 96.30% | $0.42 | 313 | 0.7 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.89% | $1.47 | 427 | 0.8 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 89.70% | $0.09 | 627 | 0.8 sec | |
| ||||||
Email triage · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 99.11% | $0.58 | 80 | 4.5 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 98.96% | $3.30 | 93 | 3.3 sec | |
| ||||||
| 3 | GPT-6 Sol OpenAIEU data | 98.44% | $2.77 | 140 | 3.3 sec | |
| ||||||
| 4 | Mistral Large 4New MistralEU data | 97.93% | $4.03 | 180 | 25.4 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 97.70% | $3.76 | 193 | 1.8 sec | |
| ||||||
| 6 | Claude Haiku 5.5New AnthropicEU data | 97.70% | $0.30 | 200 | 2.1 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 97.56% | $5.04 | 213 | 5.5 sec | |
| ||||||
| 8 | Gemma 4 31B Google · open weights | 97.11% | $0.15 | 260 | 2.2 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 96.81% | $0.16 | 280 | 3.1 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.74% | $0.46 | 280 | 11.2 sec | |
| ||||||
| 11 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.52% | $0.34 | 280 | 5.5 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 95.56% | $0.48 | 347 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.15% | $1.79 | 460 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 88.67% | $0.11 | 647 | 0.9 sec | |
| ||||||
Email triage · Spanish · Email (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 99.11% | $0.58 | 80 | 4.5 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 98.96% | $3.30 | 93 | 3.3 sec | |
| ||||||
| 3 | GPT-6 Sol OpenAIEU data | 98.44% | $2.77 | 140 | 3.3 sec | |
| ||||||
| 4 | Mistral Large 4New MistralEU data | 97.93% | $4.03 | 180 | 25.4 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 97.70% | $3.76 | 193 | 1.8 sec | |
| ||||||
| 6 | Claude Haiku 5.5New AnthropicEU data | 97.70% | $0.30 | 200 | 2.1 sec | |
| ||||||
| 7 | Claude Sonnet 5 AnthropicEU data | 97.56% | $5.04 | 213 | 5.5 sec | |
| ||||||
| 8 | Gemma 4 31B Google · open weights | 97.11% | $0.15 | 260 | 2.2 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 96.81% | $0.16 | 280 | 3.1 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.74% | $0.46 | 280 | 11.2 sec | |
| ||||||
| 11 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.52% | $0.34 | 280 | 5.5 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 95.56% | $0.48 | 347 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.15% | $1.79 | 460 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 88.67% | $0.11 | 647 | 0.9 sec | |
| ||||||
Email triage · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5 AnthropicEU data | 98.67% | $4.64 | 120 | 4.3 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 98.59% | $2.63 | 127 | 2.7 sec | |
| ||||||
| 3 | GPT-6 Sol OpenAIEU data | 98.59% | $2.81 | 127 | 3.2 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 98.52% | $0.48 | 133 | 4.2 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 98.30% | $3.70 | 153 | 24.5 sec | |
| ||||||
| 6 | Claude Sonnet 5.5New AnthropicEU data | 97.78% | $3.85 | 180 | 1.7 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 97.41% | $0.31 | 233 | 2.2 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 97.26% | $0.16 | 240 | 2.8 sec | |
| ||||||
| 9 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.74% | $0.34 | 273 | 5.2 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.44% | $0.45 | 240 | 13.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 96.22% | $0.16 | 313 | 2.4 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 96.15% | $0.49 | 313 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.15% | $1.78 | 460 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 90.52% | $0.11 | 553 | 0.9 sec | |
| ||||||
Email triage · French · Email (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5 AnthropicEU data | 98.67% | $4.64 | 120 | 4.3 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 98.59% | $2.63 | 127 | 2.7 sec | |
| ||||||
| 3 | GPT-6 Sol OpenAIEU data | 98.59% | $2.81 | 127 | 3.2 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 98.52% | $0.48 | 133 | 4.2 sec | |
| ||||||
| 5 | Mistral Large 4New MistralEU data | 98.30% | $3.70 | 153 | 24.5 sec | |
| ||||||
| 6 | Claude Sonnet 5.5New AnthropicEU data | 97.78% | $3.85 | 180 | 1.7 sec | |
| ||||||
| 7 | Claude Haiku 5.5New AnthropicEU data | 97.41% | $0.31 | 233 | 2.2 sec | |
| ||||||
| 8 | GPT-6 Luna OpenAIEU data | 97.26% | $0.16 | 240 | 2.8 sec | |
| ||||||
| 9 | DeepSeek V4.1 Flash DeepSeek · open weights | 96.74% | $0.34 | 273 | 5.2 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.44% | $0.45 | 240 | 13.2 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 96.22% | $0.16 | 313 | 2.4 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 96.15% | $0.49 | 313 | 0.8 sec | |
| ||||||
| 13 | Mistral Medium 3.5 MistralEU data | 94.15% | $1.78 | 460 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 90.52% | $0.11 | 553 | 0.9 sec | |
| ||||||
Contract clauses · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.27 | 0 | 2.6 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.40 | 0 | 2.4 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.42 | 0 | 4.3 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.69 | 0 | 4.9 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.06 | 0 | 1.2 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $3.09 | 0 | 2.2 sec | |
| ||||||
| 7 | Mistral Large 4New MistralEU data | 100.00% | $4.58 | 0 | 43.9 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $4.62 | 0 | 2.0 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.35 | 0 | 2.4 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 100.00% | $6.05 | 0 | 3.6 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 99.95% | $0.25 | 7 | 6.0 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 99.90% | $0.63 | 13 | 13.6 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 99.85% | $0.85 | 20 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 99.03% | $0.23 | 80 | 1.2 sec | |
| ||||||
Contract clauses · Czech · Contract text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.27 | 0 | 2.6 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.40 | 0 | 2.4 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.42 | 0 | 4.3 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.69 | 0 | 4.9 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.06 | 0 | 1.2 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $3.09 | 0 | 2.2 sec | |
| ||||||
| 7 | Mistral Large 4New MistralEU data | 100.00% | $4.58 | 0 | 43.9 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $4.62 | 0 | 2.0 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.35 | 0 | 2.4 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 100.00% | $6.05 | 0 | 3.6 sec | |
| ||||||
| 11 | Gemma 4 31B Google · open weights | 99.95% | $0.25 | 7 | 6.0 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 99.90% | $0.63 | 13 | 13.6 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 99.85% | $0.85 | 20 | 0.9 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 99.03% | $0.23 | 80 | 1.2 sec | |
| ||||||
Contract clauses · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.27 | 0 | 3.0 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.48 | 0 | 2.7 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.65 | 0 | 5.5 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 100.00% | $4.36 | 0 | 2.9 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $6.05 | 0 | 2.7 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.90% | $3.42 | 7 | 2.6 sec | |
| ||||||
| 7 | Mistral Large 4New MistralEU data | 99.44% | $5.12 | 47 | 36.2 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 98.21% | $7.11 | 133 | 4.3 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 98.00% | $0.78 | 127 | 19.7 sec | |
| ||||||
| 10 | Gemini 3.5 Flash Lite GoogleEU data | 98.00% | $0.79 | 160 | 1.0 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 97.13% | $2.76 | 247 | 1.2 sec | |
| ||||||
| 12 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.08% | $0.52 | 240 | 4.7 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 94.41% | $0.22 | 727 | 2.6 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 94.26% | $0.21 | 533 | 1.2 sec | |
| ||||||
Contract clauses · German · Contract text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna OpenAIEU data | 100.00% | $0.27 | 0 | 3.0 sec | |
| ||||||
| 2 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.48 | 0 | 2.7 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.65 | 0 | 5.5 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 100.00% | $4.36 | 0 | 2.9 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $6.05 | 0 | 2.7 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.90% | $3.42 | 7 | 2.6 sec | |
| ||||||
| 7 | Mistral Large 4New MistralEU data | 99.44% | $5.12 | 47 | 36.2 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 98.21% | $7.11 | 133 | 4.3 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 98.00% | $0.78 | 127 | 19.7 sec | |
| ||||||
| 10 | Gemini 3.5 Flash Lite GoogleEU data | 98.00% | $0.79 | 160 | 1.0 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 97.13% | $2.76 | 247 | 1.2 sec | |
| ||||||
| 12 | DeepSeek V4.1 Flash DeepSeek · open weights | 97.08% | $0.52 | 240 | 4.7 sec | |
| ||||||
| 13 | Gemma 4 31B Google · open weights | 94.41% | $0.22 | 727 | 2.6 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 94.26% | $0.21 | 533 | 1.2 sec | |
| ||||||
Contract clauses · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.32 | 0 | 3.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $3.33 | 0 | 2.5 sec | |
| ||||||
| 3 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.04 | 0 | 3.4 sec | |
| ||||||
| 4 | Gemini 3.5 Flash Lite GoogleEU data | 99.95% | $0.69 | 7 | 1.0 sec | |
| ||||||
| 5 | Gemini 3.8 Flash | 99.90% | $2.47 | 7 | 2.0 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.79% | $0.20 | 13 | 2.7 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 99.79% | $4.23 | 20 | 2.1 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 99.69% | $0.35 | 27 | 2.1 sec | |
| ||||||
| 9 | Gemma 4 31B Google · open weights | 99.64% | $0.18 | 27 | 2.5 sec | |
| ||||||
| 10 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $0.40 | 40 | 3.4 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 98.10% | $2.35 | 127 | 1.1 sec | |
| ||||||
| 12 | Mistral Large 4New MistralEU data | 96.10% | $5.24 | 253 | 41.4 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 95.13% | $0.62 | 307 | 14.1 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 91.54% | $0.17 | 567 | 1.1 sec | |
| ||||||
Contract clauses · English (US) · Contract text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.32 | 0 | 3.9 sec | |
| ||||||
| 2 | GPT-6 Sol OpenAIEU data | 100.00% | $3.33 | 0 | 2.5 sec | |
| ||||||
| 3 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.04 | 0 | 3.4 sec | |
| ||||||
| 4 | Gemini 3.5 Flash Lite GoogleEU data | 99.95% | $0.69 | 7 | 1.0 sec | |
| ||||||
| 5 | Gemini 3.8 Flash | 99.90% | $2.47 | 7 | 2.0 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.79% | $0.20 | 13 | 2.7 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 99.79% | $4.23 | 20 | 2.1 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 99.69% | $0.35 | 27 | 2.1 sec | |
| ||||||
| 9 | Gemma 4 31B Google · open weights | 99.64% | $0.18 | 27 | 2.5 sec | |
| ||||||
| 10 | GLM-5.3 FlashNew Z.ai · open weights | 99.49% | $0.40 | 40 | 3.4 sec | |
| ||||||
| 11 | Mistral Medium 3.5 MistralEU data | 98.10% | $2.35 | 127 | 1.1 sec | |
| ||||||
| 12 | Mistral Large 4New MistralEU data | 96.10% | $5.24 | 253 | 41.4 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 95.13% | $0.62 | 307 | 14.1 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 91.54% | $0.17 | 567 | 1.1 sec | |
| ||||||
Contract clauses · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.23 | 0 | 1.3 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.26 | 0 | 3.1 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.40 | 0 | 4.1 sec | |
| ||||||
| 4 | Qwen3.8 Flash Qwen | 100.00% | $0.60 | 0 | 14.2 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.64 | 0 | 4.9 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.02 | 0 | 1.3 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $3.33 | 0 | 3.0 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $4.48 | 0 | 2.8 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 100.00% | $4.83 | 0 | 35.0 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.70 | 0 | 2.7 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $6.58 | 0 | 4.0 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.95% | $0.42 | 7 | 2.3 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 99.79% | $0.84 | 20 | 1.1 sec | |
| ||||||
| 14 | Gemma 4 31B Google · open weights | 94.41% | $0.24 | 727 | 5.0 sec | |
| ||||||
Contract clauses · Spanish · Contract text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.23 | 0 | 1.3 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.26 | 0 | 3.1 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.40 | 0 | 4.1 sec | |
| ||||||
| 4 | Qwen3.8 Flash Qwen | 100.00% | $0.60 | 0 | 14.2 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.64 | 0 | 4.9 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.02 | 0 | 1.3 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $3.33 | 0 | 3.0 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $4.48 | 0 | 2.8 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 100.00% | $4.83 | 0 | 35.0 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.70 | 0 | 2.7 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $6.58 | 0 | 4.0 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.95% | $0.42 | 7 | 2.3 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 99.79% | $0.84 | 20 | 1.1 sec | |
| ||||||
| 14 | Gemma 4 31B Google · open weights | 94.41% | $0.24 | 727 | 5.0 sec | |
| ||||||
Contract clauses · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.01 | 0 | 1.3 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 100.00% | $3.50 | 0 | 3.0 sec | |
| ||||||
| 3 | Gemini 3.5 Flash Lite GoogleEU data | 99.95% | $0.85 | 7 | 1.0 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 99.95% | $5.71 | 7 | 2.6 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.95% | $6.69 | 7 | 4.6 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.85% | $0.45 | 20 | 5.2 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.74% | $0.69 | 33 | 5.1 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 99.59% | $4.49 | 53 | 2.6 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.49% | $0.66 | 67 | 16.2 sec | |
| ||||||
| 10 | Claude Haiku 5.5New AnthropicEU data | 99.38% | $0.45 | 80 | 2.6 sec | |
| ||||||
| 11 | GPT-6 Luna OpenAIEU data | 99.33% | $0.27 | 87 | 3.2 sec | |
| ||||||
| 12 | Mistral Large 4New MistralEU data | 99.28% | $4.60 | 93 | 34.7 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.23% | $0.23 | 100 | 1.2 sec | |
| ||||||
| 14 | Gemma 4 31B Google · open weights | 94.62% | $0.26 | 693 | 5.3 sec | |
| ||||||
Contract clauses · French · Contract text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Medium 3.5 MistralEU data | 100.00% | $3.01 | 0 | 1.3 sec | |
| ||||||
| 2 | Gemini 3.8 Flash | 100.00% | $3.50 | 0 | 3.0 sec | |
| ||||||
| 3 | Gemini 3.5 Flash Lite GoogleEU data | 99.95% | $0.85 | 7 | 1.0 sec | |
| ||||||
| 4 | Claude Sonnet 5.5New AnthropicEU data | 99.95% | $5.71 | 7 | 2.6 sec | |
| ||||||
| 5 | Claude Sonnet 5 AnthropicEU data | 99.95% | $6.69 | 7 | 4.6 sec | |
| ||||||
| 6 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.85% | $0.45 | 20 | 5.2 sec | |
| ||||||
| 7 | GLM-5.3 FlashNew Z.ai · open weights | 99.74% | $0.69 | 33 | 5.1 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 99.59% | $4.49 | 53 | 2.6 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.49% | $0.66 | 67 | 16.2 sec | |
| ||||||
| 10 | Claude Haiku 5.5New AnthropicEU data | 99.38% | $0.45 | 80 | 2.6 sec | |
| ||||||
| 11 | GPT-6 Luna OpenAIEU data | 99.33% | $0.27 | 87 | 3.2 sec | |
| ||||||
| 12 | Mistral Large 4New MistralEU data | 99.28% | $4.60 | 93 | 34.7 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.23% | $0.23 | 100 | 1.2 sec | |
| ||||||
| 14 | Gemma 4 31B Google · open weights | 94.62% | $0.26 | 693 | 5.3 sec | |
| ||||||
Personal data detection · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 99.81% | $0.12 | 13 | 2.9 sec | |
| ||||||
| 2 | Mistral Medium 3.5 MistralEU data | 99.81% | $1.40 | 13 | 0.8 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.71% | $0.31 | 20 | 4.2 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.71% | $2.22 | 20 | 2.5 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 99.71% | $3.52 | 20 | 2.2 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 1.8 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 99.52% | $2.19 | 33 | 2.0 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 99.43% | $0.21 | 40 | 1.8 sec | |
| ||||||
| 9 | GLM-5.3 FlashNew Z.ai · open weights | 99.33% | $0.49 | 7 | 3.8 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 99.33% | $3.15 | 47 | 18.5 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 98.86% | $4.06 | 80 | 3.2 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 98.76% | $0.09 | 80 | 0.9 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 98.10% | $0.45 | 120 | 0.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 96.95% | $0.41 | 113 | 9.6 sec | |
| ||||||
Personal data detection · Czech · Text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 99.81% | $0.12 | 13 | 2.9 sec | |
| ||||||
| 2 | Mistral Medium 3.5 MistralEU data | 99.81% | $1.40 | 13 | 0.8 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.71% | $0.31 | 20 | 4.2 sec | |
| ||||||
| 4 | GPT-6 Sol OpenAIEU data | 99.71% | $2.22 | 20 | 2.5 sec | |
| ||||||
| 5 | Claude Sonnet 5.5New AnthropicEU data | 99.71% | $3.52 | 20 | 2.2 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 1.8 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 99.52% | $2.19 | 33 | 2.0 sec | |
| ||||||
| 8 | Claude Haiku 5.5New AnthropicEU data | 99.43% | $0.21 | 40 | 1.8 sec | |
| ||||||
| 9 | GLM-5.3 FlashNew Z.ai · open weights | 99.33% | $0.49 | 7 | 3.8 sec | |
| ||||||
| 10 | Mistral Large 4New MistralEU data | 99.33% | $3.15 | 47 | 18.5 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 98.86% | $4.06 | 80 | 3.2 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 98.76% | $0.09 | 80 | 0.9 sec | |
| ||||||
| 13 | Gemini 3.5 Flash Lite GoogleEU data | 98.10% | $0.45 | 120 | 0.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 96.95% | $0.41 | 113 | 9.6 sec | |
| ||||||
Personal data detection · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.38 | 0 | 0.9 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.87 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.10 | 7 | 1.9 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.81% | $0.31 | 13 | 3.7 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.40 | 13 | 3.5 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.71% | $2.12 | 20 | 2.3 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.13 | 20 | 2.6 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $4.40 | 20 | 3.7 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.62% | $0.36 | 7 | 12.2 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 2.2 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 99.62% | $0.23 | 27 | 1.8 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.62% | $0.45 | 27 | 0.8 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 99.52% | $2.61 | 33 | 18.6 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 99.14% | $0.09 | 60 | 0.9 sec | |
| ||||||
Personal data detection · German · Text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.38 | 0 | 0.9 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.87 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.10 | 7 | 1.9 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.81% | $0.31 | 13 | 3.7 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.40 | 13 | 3.5 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.71% | $2.12 | 20 | 2.3 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.13 | 20 | 2.6 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $4.40 | 20 | 3.7 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.62% | $0.36 | 7 | 12.2 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 2.2 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 99.62% | $0.23 | 27 | 1.8 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.62% | $0.45 | 27 | 0.8 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 99.52% | $2.61 | 33 | 18.6 sec | |
| ||||||
| 14 | Mistral Small 4 Mistral · open weightsEU data | 99.14% | $0.09 | 60 | 0.9 sec | |
| ||||||
Personal data detection · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.33 | 0 | 2.8 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.22 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.08 | 7 | 2.4 sec | |
| ||||||
| 4 | Claude Haiku 5.5New AnthropicEU data | 99.81% | $0.19 | 13 | 1.7 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 99.71% | $1.15 | 20 | 0.8 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.71% | $1.86 | 20 | 1.9 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.09 | 20 | 2.6 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $3.50 | 20 | 2.5 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 99.62% | $0.11 | 27 | 1.8 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.62% | $0.29 | 27 | 3.8 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 99.62% | $2.89 | 27 | 28.2 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.52% | $0.41 | 27 | 0.8 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.43% | $0.08 | 40 | 0.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 99.24% | $0.28 | 13 | 7.5 sec | |
| ||||||
Personal data detection · English (US) · Text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.33 | 0 | 2.8 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.22 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.08 | 7 | 2.4 sec | |
| ||||||
| 4 | Claude Haiku 5.5New AnthropicEU data | 99.81% | $0.19 | 13 | 1.7 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 99.71% | $1.15 | 20 | 0.8 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 99.71% | $1.86 | 20 | 1.9 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.09 | 20 | 2.6 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $3.50 | 20 | 2.5 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 99.62% | $0.11 | 27 | 1.8 sec | |
| ||||||
| 10 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.62% | $0.29 | 27 | 3.8 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 99.62% | $2.89 | 27 | 28.2 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.52% | $0.41 | 27 | 0.8 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.43% | $0.08 | 40 | 0.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 99.24% | $0.28 | 13 | 7.5 sec | |
| ||||||
Personal data detection · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.23 | 0 | 1.9 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.78 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.10 | 7 | 3.6 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.90% | $0.44 | 7 | 5.9 sec | |
| ||||||
| 5 | Qwen3.8 Flash Qwen | 99.81% | $0.38 | 13 | 10.2 sec | |
| ||||||
| 6 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.46 | 13 | 3.4 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 99.81% | $2.11 | 13 | 2.3 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 99.81% | $2.21 | 13 | 2.9 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 99.81% | $3.16 | 13 | 21.9 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 99.81% | $4.38 | 13 | 3.6 sec | |
| ||||||
| 11 | GPT-6 Luna OpenAIEU data | 99.71% | $0.12 | 20 | 2.3 sec | |
| ||||||
| 12 | Mistral Medium 3.5 MistralEU data | 99.62% | $1.41 | 27 | 1.0 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.05% | $0.09 | 67 | 0.9 sec | |
| ||||||
| 14 | Gemini 3.5 Flash Lite GoogleEU data | 95.81% | $0.45 | 267 | 0.9 sec | |
| ||||||
Personal data detection · Spanish · Text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.23 | 0 | 1.9 sec | |
| ||||||
| 2 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.78 | 0 | 1.9 sec | |
| ||||||
| 3 | Gemma 4 31B Google · open weights | 99.90% | $0.10 | 7 | 3.6 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.90% | $0.44 | 7 | 5.9 sec | |
| ||||||
| 5 | Qwen3.8 Flash Qwen | 99.81% | $0.38 | 13 | 10.2 sec | |
| ||||||
| 6 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.46 | 13 | 3.4 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 99.81% | $2.11 | 13 | 2.3 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 99.81% | $2.21 | 13 | 2.9 sec | |
| ||||||
| 9 | Mistral Large 4New MistralEU data | 99.81% | $3.16 | 13 | 21.9 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 99.81% | $4.38 | 13 | 3.6 sec | |
| ||||||
| 11 | GPT-6 Luna OpenAIEU data | 99.71% | $0.12 | 20 | 2.3 sec | |
| ||||||
| 12 | Mistral Medium 3.5 MistralEU data | 99.62% | $1.41 | 27 | 1.0 sec | |
| ||||||
| 13 | Mistral Small 4 Mistral · open weightsEU data | 99.05% | $0.09 | 67 | 0.9 sec | |
| ||||||
| 14 | Gemini 3.5 Flash Lite GoogleEU data | 95.81% | $0.45 | 267 | 0.9 sec | |
| ||||||
Personal data detection · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.81 | 0 | 2.0 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 99.81% | $0.12 | 13 | 4.3 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.41 | 13 | 3.3 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 99.81% | $2.13 | 13 | 2.3 sec | |
| ||||||
| 5 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.71% | $0.37 | 20 | 5.3 sec | |
| ||||||
| 6 | Qwen3.8 Flash Qwen | 99.71% | $0.40 | 20 | 10.9 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.23 | 20 | 2.9 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $4.10 | 20 | 2.5 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 2.4 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 99.62% | $1.43 | 27 | 0.9 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 99.33% | $0.23 | 47 | 1.8 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 98.95% | $0.09 | 67 | 1.0 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 98.67% | $2.96 | 53 | 23.6 sec | |
| ||||||
| 14 | Gemini 3.5 Flash Lite GoogleEU data | 94.95% | $0.46 | 327 | 0.8 sec | |
| ||||||
Personal data detection · French · Text (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.81 | 0 | 2.0 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 99.81% | $0.12 | 13 | 4.3 sec | |
| ||||||
| 3 | GLM-5.3 FlashNew Z.ai · open weights | 99.81% | $0.41 | 13 | 3.3 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 99.81% | $2.13 | 13 | 2.3 sec | |
| ||||||
| 5 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.71% | $0.37 | 20 | 5.3 sec | |
| ||||||
| 6 | Qwen3.8 Flash Qwen | 99.71% | $0.40 | 20 | 10.9 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 99.71% | $2.23 | 20 | 2.9 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 99.71% | $4.10 | 20 | 2.5 sec | |
| ||||||
| 9 | GPT-6 Luna OpenAIEU data | 99.62% | $0.12 | 27 | 2.4 sec | |
| ||||||
| 10 | Mistral Medium 3.5 MistralEU data | 99.62% | $1.43 | 27 | 0.9 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 99.33% | $0.23 | 47 | 1.8 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 98.95% | $0.09 | 67 | 1.0 sec | |
| ||||||
| 13 | Mistral Large 4New MistralEU data | 98.67% | $2.96 | 53 | 23.6 sec | |
| ||||||
| 14 | Gemini 3.5 Flash Lite GoogleEU data | 94.95% | $0.46 | 327 | 0.8 sec | |
| ||||||
Document summaries · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.19 | 0 | 3.4 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.21 | 0 | 3.4 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.78 | 0 | 13.4 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.13 | 0 | 10.4 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.31 | 0 | 1.4 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 100.00% | $4.72 | 0 | 7.6 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $6.69 | 0 | 6.2 sec | |
| ||||||
| 8 | Gemini 3.5 Flash Lite GoogleEU data | 99.56% | $0.69 | 27 | 1.0 sec | |
| ||||||
| 9 | Mistral Small 4 Mistral · open weightsEU data | 99.00% | $0.22 | 60 | 1.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 98.67% | $5.01 | 80 | 2.7 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 95.89% | $0.32 | 227 | 2.1 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 95.44% | $1.48 | 67 | 34.2 sec | |
| ||||||
| 13 | Claude Sonnet 5 AnthropicEU data | 93.44% | $6.48 | 320 | 4.7 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 83.33% | $10.72 | 167 | 55.4 sec | |
| ||||||
Document summaries · Czech · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.19 | 0 | 3.4 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.21 | 0 | 3.4 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.78 | 0 | 13.4 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.13 | 0 | 10.4 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.31 | 0 | 1.4 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 100.00% | $4.72 | 0 | 7.6 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $6.69 | 0 | 6.2 sec | |
| ||||||
| 8 | Gemini 3.5 Flash Lite GoogleEU data | 99.56% | $0.69 | 27 | 1.0 sec | |
| ||||||
| 9 | Mistral Small 4 Mistral · open weightsEU data | 99.00% | $0.22 | 60 | 1.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 98.67% | $5.01 | 80 | 2.7 sec | |
| ||||||
| 11 | Claude Haiku 5.5New AnthropicEU data | 95.89% | $0.32 | 227 | 2.1 sec | |
| ||||||
| 12 | Qwen3.8 Flash Qwen | 95.44% | $1.48 | 67 | 34.2 sec | |
| ||||||
| 13 | Claude Sonnet 5 AnthropicEU data | 93.44% | $6.48 | 320 | 4.7 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 83.33% | $10.72 | 167 | 55.4 sec | |
| ||||||
Document summaries · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.16 | 0 | 2.9 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.19 | 0 | 3.3 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.62 | 0 | 9.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.08 | 0 | 8.4 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.98 | 0 | 1.1 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 100.00% | $4.11 | 0 | 7.1 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $6.11 | 0 | 5.8 sec | |
| ||||||
| 8 | Mistral Small 4 Mistral · open weightsEU data | 99.89% | $0.19 | 7 | 1.3 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 99.89% | $0.59 | 7 | 0.9 sec | |
| ||||||
| 10 | Claude Haiku 5.5New AnthropicEU data | 99.44% | $0.46 | 33 | 3.5 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 99.33% | $7.19 | 7 | 48.0 sec | |
| ||||||
| 12 | Claude Sonnet 5 AnthropicEU data | 99.11% | $5.97 | 40 | 3.6 sec | |
| ||||||
| 13 | Claude Sonnet 5.5New AnthropicEU data | 99.11% | $5.82 | 53 | 2.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 95.33% | $1.43 | 47 | 29.5 sec | |
| ||||||
Document summaries · German · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.16 | 0 | 2.9 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.19 | 0 | 3.3 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.62 | 0 | 9.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.08 | 0 | 8.4 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.98 | 0 | 1.1 sec | |
| ||||||
| 6 | GPT-6 Sol OpenAIEU data | 100.00% | $4.11 | 0 | 7.1 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $6.11 | 0 | 5.8 sec | |
| ||||||
| 8 | Mistral Small 4 Mistral · open weightsEU data | 99.89% | $0.19 | 7 | 1.3 sec | |
| ||||||
| 9 | Gemini 3.5 Flash Lite GoogleEU data | 99.89% | $0.59 | 7 | 0.9 sec | |
| ||||||
| 10 | Claude Haiku 5.5New AnthropicEU data | 99.44% | $0.46 | 33 | 3.5 sec | |
| ||||||
| 11 | Mistral Large 4New MistralEU data | 99.33% | $7.19 | 7 | 48.0 sec | |
| ||||||
| 12 | Claude Sonnet 5 AnthropicEU data | 99.11% | $5.97 | 40 | 3.6 sec | |
| ||||||
| 13 | Claude Sonnet 5.5New AnthropicEU data | 99.11% | $5.82 | 53 | 2.8 sec | |
| ||||||
| 14 | Qwen3.8 Flash Qwen | 95.33% | $1.43 | 47 | 29.5 sec | |
| ||||||
Document summaries · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.13 | 0 | 2.2 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.16 | 0 | 2.9 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.54 | 0 | 8.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.82 | 0 | 7.5 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.57 | 0 | 0.9 sec | |
| ||||||
| 6 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.60 | 0 | 1.8 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $5.46 | 0 | 5.4 sec | |
| ||||||
| 8 | Mistral Large 4New MistralEU data | 100.00% | $5.62 | 0 | 37.2 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.89% | $0.75 | 7 | 17.8 sec | |
| ||||||
| 10 | Mistral Small 4 Mistral · open weightsEU data | 99.67% | $0.14 | 13 | 1.0 sec | |
| ||||||
| 11 | GPT-6 Sol OpenAIEU data | 99.56% | $3.85 | 27 | 8.1 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.22% | $0.46 | 47 | 0.9 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 98.33% | $0.34 | 100 | 2.7 sec | |
| ||||||
| 14 | Claude Sonnet 5 AnthropicEU data | 96.00% | $4.08 | 233 | 2.9 sec | |
| ||||||
Document summaries · English (US) · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.13 | 0 | 2.2 sec | |
| ||||||
| 2 | GPT-6 Luna OpenAIEU data | 100.00% | $0.16 | 0 | 2.9 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.54 | 0 | 8.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.82 | 0 | 7.5 sec | |
| ||||||
| 5 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.57 | 0 | 0.9 sec | |
| ||||||
| 6 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.60 | 0 | 1.8 sec | |
| ||||||
| 7 | Gemini 3.8 Flash | 100.00% | $5.46 | 0 | 5.4 sec | |
| ||||||
| 8 | Mistral Large 4New MistralEU data | 100.00% | $5.62 | 0 | 37.2 sec | |
| ||||||
| 9 | Qwen3.8 Flash Qwen | 99.89% | $0.75 | 7 | 17.8 sec | |
| ||||||
| 10 | Mistral Small 4 Mistral · open weightsEU data | 99.67% | $0.14 | 13 | 1.0 sec | |
| ||||||
| 11 | GPT-6 Sol OpenAIEU data | 99.56% | $3.85 | 27 | 8.1 sec | |
| ||||||
| 12 | Gemini 3.5 Flash Lite GoogleEU data | 99.22% | $0.46 | 47 | 0.9 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 98.33% | $0.34 | 100 | 2.7 sec | |
| ||||||
| 14 | Claude Sonnet 5 AnthropicEU data | 96.00% | $4.08 | 233 | 2.9 sec | |
| ||||||
Document summaries · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.88 | 0 | 10.6 sec | |
| ||||||
| 2 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.86 | 0 | 1.0 sec | |
| ||||||
| 3 | Gemini 3.8 Flash | 100.00% | $6.78 | 0 | 5.9 sec | |
| ||||||
| 4 | Gemma 4 31B Google · open weights | 99.67% | $0.16 | 20 | 2.6 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 97.89% | $5.02 | 127 | 7.0 sec | |
| ||||||
| 6 | Mistral Small 4 Mistral · open weightsEU data | 97.78% | $0.18 | 133 | 1.2 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 97.00% | $0.19 | 180 | 3.0 sec | |
| ||||||
| 8 | GLM-5.3 FlashNew Z.ai · open weights | 96.67% | $1.54 | 33 | 9.2 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 95.56% | $5.55 | 267 | 2.1 sec | |
| ||||||
| 10 | Gemini 3.5 Flash Lite GoogleEU data | 95.33% | $0.56 | 280 | 0.9 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 94.44% | $1.22 | 67 | 23.9 sec | |
| ||||||
| 12 | Claude Sonnet 5 AnthropicEU data | 94.22% | $6.32 | 320 | 3.4 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 94.11% | $0.61 | 353 | 3.8 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 91.33% | $7.99 | 87 | 38.3 sec | |
| ||||||
Document summaries · Spanish · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.88 | 0 | 10.6 sec | |
| ||||||
| 2 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.86 | 0 | 1.0 sec | |
| ||||||
| 3 | Gemini 3.8 Flash | 100.00% | $6.78 | 0 | 5.9 sec | |
| ||||||
| 4 | Gemma 4 31B Google · open weights | 99.67% | $0.16 | 20 | 2.6 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 97.89% | $5.02 | 127 | 7.0 sec | |
| ||||||
| 6 | Mistral Small 4 Mistral · open weightsEU data | 97.78% | $0.18 | 133 | 1.2 sec | |
| ||||||
| 7 | GPT-6 Luna OpenAIEU data | 97.00% | $0.19 | 180 | 3.0 sec | |
| ||||||
| 8 | GLM-5.3 FlashNew Z.ai · open weights | 96.67% | $1.54 | 33 | 9.2 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 95.56% | $5.55 | 267 | 2.1 sec | |
| ||||||
| 10 | Gemini 3.5 Flash Lite GoogleEU data | 95.33% | $0.56 | 280 | 0.9 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 94.44% | $1.22 | 67 | 23.9 sec | |
| ||||||
| 12 | Claude Sonnet 5 AnthropicEU data | 94.22% | $6.32 | 320 | 3.4 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 94.11% | $0.61 | 353 | 3.8 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 91.33% | $7.99 | 87 | 38.3 sec | |
| ||||||
Document summaries · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.78 | 0 | 10.5 sec | |
| ||||||
| 2 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.25 | 0 | 8.7 sec | |
| ||||||
| 3 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.87 | 0 | 1.0 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 100.00% | $7.06 | 0 | 6.7 sec | |
| ||||||
| 5 | Gemma 4 31B Google · open weights | 99.11% | $0.16 | 53 | 3.2 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 98.78% | $0.20 | 73 | 3.2 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 97.89% | $4.89 | 127 | 2.7 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 97.22% | $4.58 | 167 | 6.1 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 96.78% | $5.04 | 193 | 3.1 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.33% | $1.46 | 53 | 28.1 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 94.67% | $0.65 | 287 | 0.9 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 93.22% | $0.19 | 380 | 1.3 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 92.78% | $0.66 | 433 | 4.0 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 72.67% | $11.26 | 273 | 46.4 sec | |
| ||||||
Document summaries · French · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.78 | 0 | 10.5 sec | |
| ||||||
| 2 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $1.25 | 0 | 8.7 sec | |
| ||||||
| 3 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.87 | 0 | 1.0 sec | |
| ||||||
| 4 | Gemini 3.8 Flash | 100.00% | $7.06 | 0 | 6.7 sec | |
| ||||||
| 5 | Gemma 4 31B Google · open weights | 99.11% | $0.16 | 53 | 3.2 sec | |
| ||||||
| 6 | GPT-6 Luna OpenAIEU data | 98.78% | $0.20 | 73 | 3.2 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 97.89% | $4.89 | 127 | 2.7 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 97.22% | $4.58 | 167 | 6.1 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 96.78% | $5.04 | 193 | 3.1 sec | |
| ||||||
| 10 | Qwen3.8 Flash Qwen | 96.33% | $1.46 | 53 | 28.1 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 94.67% | $0.65 | 287 | 0.9 sec | |
| ||||||
| 12 | Mistral Small 4 Mistral · open weightsEU data | 93.22% | $0.19 | 380 | 1.3 sec | |
| ||||||
| 13 | Claude Haiku 5.5New AnthropicEU data | 92.78% | $0.66 | 433 | 4.0 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 72.67% | $11.26 | 273 | 46.4 sec | |
| ||||||
Questions about a document · Czech · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.21 | 0 | 1.9 sec | |
| ||||||
| 2 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.21 | 0 | 0.8 sec | |
| ||||||
| 3 | GPT-6 Luna OpenAIEU data | 100.00% | $0.23 | 0 | 3.1 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.38 | 0 | 3.5 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.46 | 0 | 3.0 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.58 | 0 | 0.8 sec | |
| ||||||
| 7 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.51 | 0 | 0.9 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $3.95 | 0 | 2.3 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 100.00% | $4.36 | 0 | 3.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.03 | 0 | 2.1 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.14 | 0 | 2.6 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.78% | $0.30 | 13 | 1.9 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 94.22% | $0.71 | 153 | 14.3 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 89.89% | $7.08 | 220 | 33.6 sec | |
| ||||||
Questions about a document · Czech · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.21 | 0 | 1.9 sec | |
| ||||||
| 2 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.21 | 0 | 0.8 sec | |
| ||||||
| 3 | GPT-6 Luna OpenAIEU data | 100.00% | $0.23 | 0 | 3.1 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.38 | 0 | 3.5 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.46 | 0 | 3.0 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.58 | 0 | 0.8 sec | |
| ||||||
| 7 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.51 | 0 | 0.9 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $3.95 | 0 | 2.3 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 100.00% | $4.36 | 0 | 3.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.03 | 0 | 2.1 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.14 | 0 | 2.6 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.78% | $0.30 | 13 | 1.9 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 94.22% | $0.71 | 153 | 14.3 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 89.89% | $7.08 | 220 | 33.6 sec | |
| ||||||
Questions about a document · German · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.18 | 0 | 2.1 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.32 | 0 | 1.9 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.33 | 0 | 3.0 sec | |
| ||||||
| 5 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.52 | 0 | 0.7 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.13 | 0 | 0.8 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $3.23 | 0 | 2.2 sec | |
| ||||||
| 8 | Gemini 3.8 Flash | 100.00% | $3.76 | 0 | 3.2 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.49 | 0 | 1.9 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.49 | 0 | 2.3 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 98.89% | $0.46 | 33 | 8.6 sec | |
| ||||||
| 12 | GLM-5.3 FlashNew Z.ai · open weights | 98.78% | $0.41 | 47 | 2.3 sec | |
| ||||||
| 13 | GPT-6 Luna OpenAIEU data | 98.00% | $0.20 | 87 | 3.2 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 93.56% | $5.77 | 107 | 31.5 sec | |
| ||||||
Questions about a document · German · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.18 | 0 | 2.1 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.32 | 0 | 1.9 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.33 | 0 | 3.0 sec | |
| ||||||
| 5 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.52 | 0 | 0.7 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.13 | 0 | 0.8 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $3.23 | 0 | 2.2 sec | |
| ||||||
| 8 | Gemini 3.8 Flash | 100.00% | $3.76 | 0 | 3.2 sec | |
| ||||||
| 9 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $5.49 | 0 | 1.9 sec | |
| ||||||
| 10 | Claude Sonnet 5 AnthropicEU data | 100.00% | $5.49 | 0 | 2.3 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 98.89% | $0.46 | 33 | 8.6 sec | |
| ||||||
| 12 | GLM-5.3 FlashNew Z.ai · open weights | 98.78% | $0.41 | 47 | 2.3 sec | |
| ||||||
| 13 | GPT-6 Luna OpenAIEU data | 98.00% | $0.20 | 87 | 3.2 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 93.56% | $5.77 | 107 | 31.5 sec | |
| ||||||
Questions about a document · English (US) · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.15 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.15 | 0 | 2.1 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.23 | 0 | 1.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.32 | 0 | 2.4 sec | |
| ||||||
| 5 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.46 | 0 | 0.8 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.81 | 0 | 0.8 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $2.36 | 0 | 2.3 sec | |
| ||||||
| 8 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.84 | 0 | 1.5 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 100.00% | $3.88 | 0 | 2.2 sec | |
| ||||||
| 10 | Gemini 3.8 Flash | 99.89% | $3.00 | 7 | 3.0 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 99.78% | $0.44 | 13 | 8.7 sec | |
| ||||||
| 12 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.22% | $0.31 | 13 | 4.2 sec | |
| ||||||
| 13 | GPT-6 Luna OpenAIEU data | 99.11% | $0.15 | 40 | 3.1 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 98.89% | $4.16 | 27 | 17.7 sec | |
| ||||||
Questions about a document · English (US) · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.15 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.15 | 0 | 2.1 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.23 | 0 | 1.8 sec | |
| ||||||
| 4 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.32 | 0 | 2.4 sec | |
| ||||||
| 5 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.46 | 0 | 0.8 sec | |
| ||||||
| 6 | Mistral Medium 3.5 MistralEU data | 100.00% | $1.81 | 0 | 0.8 sec | |
| ||||||
| 7 | GPT-6 Sol OpenAIEU data | 100.00% | $2.36 | 0 | 2.3 sec | |
| ||||||
| 8 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $3.84 | 0 | 1.5 sec | |
| ||||||
| 9 | Claude Sonnet 5 AnthropicEU data | 100.00% | $3.88 | 0 | 2.2 sec | |
| ||||||
| 10 | Gemini 3.8 Flash | 99.89% | $3.00 | 7 | 3.0 sec | |
| ||||||
| 11 | Qwen3.8 Flash Qwen | 99.78% | $0.44 | 13 | 8.7 sec | |
| ||||||
| 12 | DeepSeek V4.1 Flash DeepSeek · open weights | 99.22% | $0.31 | 13 | 4.2 sec | |
| ||||||
| 13 | GPT-6 Luna OpenAIEU data | 99.11% | $0.15 | 40 | 3.1 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 98.89% | $4.16 | 27 | 17.7 sec | |
| ||||||
Questions about a document · Spanish · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.18 | 0 | 1.9 sec | |
| ||||||
| 2 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.29 | 0 | 1.8 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.35 | 0 | 3.3 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.36 | 0 | 2.3 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.51 | 0 | 0.8 sec | |
| ||||||
| 7 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.18 | 0 | 0.8 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $3.36 | 0 | 2.5 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 100.00% | $4.05 | 0 | 3.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $4.89 | 0 | 1.4 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $4.94 | 0 | 2.2 sec | |
| ||||||
| 12 | GPT-6 Luna OpenAIEU data | 99.67% | $0.19 | 13 | 2.9 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 99.67% | $0.53 | 13 | 10.3 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 94.22% | $5.74 | 93 | 22.8 sec | |
| ||||||
Questions about a document · Spanish · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 31B Google · open weights | 100.00% | $0.18 | 0 | 1.9 sec | |
| ||||||
| 2 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 3 | Claude Haiku 5.5New AnthropicEU data | 100.00% | $0.29 | 0 | 1.8 sec | |
| ||||||
| 4 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.35 | 0 | 3.3 sec | |
| ||||||
| 5 | GLM-5.3 FlashNew Z.ai · open weights | 100.00% | $0.36 | 0 | 2.3 sec | |
| ||||||
| 6 | Gemini 3.5 Flash Lite GoogleEU data | 100.00% | $0.51 | 0 | 0.8 sec | |
| ||||||
| 7 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.18 | 0 | 0.8 sec | |
| ||||||
| 8 | GPT-6 Sol OpenAIEU data | 100.00% | $3.36 | 0 | 2.5 sec | |
| ||||||
| 9 | Gemini 3.8 Flash | 100.00% | $4.05 | 0 | 3.6 sec | |
| ||||||
| 10 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $4.89 | 0 | 1.4 sec | |
| ||||||
| 11 | Claude Sonnet 5 AnthropicEU data | 100.00% | $4.94 | 0 | 2.2 sec | |
| ||||||
| 12 | GPT-6 Luna OpenAIEU data | 99.67% | $0.19 | 13 | 2.9 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 99.67% | $0.53 | 13 | 10.3 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 94.22% | $5.74 | 93 | 22.8 sec | |
| ||||||
Questions about a document · French · All levels (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.19 | 0 | 2.1 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.36 | 0 | 3.6 sec | |
| ||||||
| 4 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.19 | 0 | 0.9 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $3.43 | 0 | 2.5 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $4.06 | 0 | 3.7 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $4.90 | 0 | 1.3 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 100.00% | $4.93 | 0 | 2.1 sec | |
| ||||||
| 9 | GLM-5.3 FlashNew Z.ai · open weights | 99.67% | $0.39 | 13 | 2.2 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 99.56% | $0.20 | 20 | 3.0 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 99.56% | $0.53 | 20 | 0.8 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.22% | $0.28 | 33 | 1.8 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 98.78% | $0.57 | 27 | 10.4 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 92.33% | $5.74 | 127 | 19.8 sec | |
| ||||||
Questions about a document · French · Document (150 documents)
Best accuracy for the price meets below
| # | Model | Correct fieldsThe bar: the share of correct fields below which manual fixes cost more than a cheaper model saves. The black mark on the meter shows it. | Per 1,000 documents | To fix / 1,000How many of 1,000 documents have at least one wrong field and need someone to check and fix them. | Speed (median) | Details |
|---|---|---|---|---|---|---|
| 1 | Mistral Small 4 Mistral · open weightsEU data | 100.00% | $0.18 | 0 | 0.8 sec | |
| ||||||
| 2 | Gemma 4 31B Google · open weights | 100.00% | $0.19 | 0 | 2.1 sec | |
| ||||||
| 3 | DeepSeek V4.1 Flash DeepSeek · open weights | 100.00% | $0.36 | 0 | 3.6 sec | |
| ||||||
| 4 | Mistral Medium 3.5 MistralEU data | 100.00% | $2.19 | 0 | 0.9 sec | |
| ||||||
| 5 | GPT-6 Sol OpenAIEU data | 100.00% | $3.43 | 0 | 2.5 sec | |
| ||||||
| 6 | Gemini 3.8 Flash | 100.00% | $4.06 | 0 | 3.7 sec | |
| ||||||
| 7 | Claude Sonnet 5.5New AnthropicEU data | 100.00% | $4.90 | 0 | 1.3 sec | |
| ||||||
| 8 | Claude Sonnet 5 AnthropicEU data | 100.00% | $4.93 | 0 | 2.1 sec | |
| ||||||
| 9 | GLM-5.3 FlashNew Z.ai · open weights | 99.67% | $0.39 | 13 | 2.2 sec | |
| ||||||
| 10 | GPT-6 Luna OpenAIEU data | 99.56% | $0.20 | 20 | 3.0 sec | |
| ||||||
| 11 | Gemini 3.5 Flash Lite GoogleEU data | 99.56% | $0.53 | 20 | 0.8 sec | |
| ||||||
| 12 | Claude Haiku 5.5New AnthropicEU data | 99.22% | $0.28 | 33 | 1.8 sec | |
| ||||||
| 13 | Qwen3.8 Flash Qwen | 98.78% | $0.57 | 27 | 10.4 sec | |
| ||||||
| 14 | Mistral Large 4New MistralEU data | 92.33% | $5.74 | 127 | 19.8 sec | |
| ||||||