ModelVerdict

Sarah Villanueva · 2025-7876

English (US)Clean image11 fields

Document en-p0025 · layout en-L01 · 11 fields · public part of the dataset · Same invoice in Czech → · Report an error in this example

Sarah Villanueva · 2025-7876

Verdict

Cheapest without an error: Mistral Small 4 at $0.49 per 1,000 documents.

  • GPT-6 Sol is also error-free but 28.2× more expensive.
  • Gemma 4 31B is 3.4× cheaper, but: wrong value (Bank account).

Prompt given to every model

Extract the following data from the attached document and return only JSON with these keys:

supplier_name      supplier's name (exactly as printed, including the legal form)
supplier_id        supplier's tax ID (EIN), if printed
invoice_number     document number
payment_reference  payment reference to quote with the payment
issue_date         invoice date (YYYY-MM-DD)
due_date           due date (YYYY-MM-DD)
net_by_rate        amount per sales tax rate, e.g. {"8.25": 1845.00}; non-taxable amounts under "0"
vat_by_rate        sales tax per rate, e.g. {"8.25": 152.21}
total              total due (number, negative for credit notes)
currency           ISO 4217 currency code (USD, GBP, EUR, …)
bank_account       supplier's bank details as "routing-number account-number", or IBAN

Rules:
- Return the supplier's data only, never the customer's.
- If a value is not on the document, return null. Do not guess.
- If the document has no tax breakdown, return net_by_rate and vat_by_rate as null.
- Amounts as numbers with a decimal point, no thousands separators and no currency symbol.

How we score

Exact match per field after normalising dates, amounts and whitespace. Missing diacritics count as an error. Cost is the real token spend of this run, converted at the rate shown in the footer.

What each model extracted

FieldCorrect valueGemma 4 31B10 / 11 fieldsMistral Small 411 / 11 fieldsDeepSeek V4.1 Flash11 / 11 fieldsQwen3.8 Flash11 / 11 fieldsGPT-6 Luna11 / 11 fieldsGemini 3.5 Flash Lite11 / 11 fieldsGemini 3.8 Flash11 / 11 fieldsMistral Medium 3.511 / 11 fieldsClaude Sonnet 511 / 11 fieldsGPT-6 Sol11 / 11 fields
SupplierSarah VillanuevaSarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)Sarah Villanueva (correct)
Supplier ID (EIN, IČO…)—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Invoice number2025-78762025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)
Payment reference2025-78762025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)2025–7876 (correct)2025-7876 (correct)2025–7876 (correct)2025-7876 (correct)2025-7876 (correct)2025-7876 (correct)
Invoice date2025-09-132025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)2025-09-13 (correct)
Due date2025-10-132025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)2025-10-13 (correct)
Net by tax rate—0 %: 19638.03 (correct)0 %: 19638.03 (correct)— (correct)0 %: 19638.03 (correct)— (correct)0 %: 19638.03 (correct)0 %: 19638.03 (correct)0 %: 19638.03 (correct)0 %: 19638.03 (correct)— (correct)
Tax by rate—— (correct)— (correct)— (correct)0 %: 0.00 (correct)— (correct)0 %: 0.00 (correct)0 %: 0.00 (correct)— (correct)— (correct)— (correct)
Total19638.0319638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)19638.03 (correct)
CurrencyUSDUSD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)
Bank account043000096 2911246075 043000896 2911246875wrong value043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)043000096 2911246075 (correct)
Tokensinput / output584 / 1582,693 / 1501,549 / 4623,108 / 7134,871 / 4051,406 / 1981,406 / 9732,693 / 1495,163 / 1764,871 / 173
of which hidden reasoningbilled as output0035255230808320076
of which from cacheinput billed at a discount3070436384000000
Cost of this documentthis run$0.0001$0.0005$0.0006$0.0007$0.0008$0.0009$0.0047$0.0052$0.0121$0.0139
Per 1,000 documentsfor the same kind of document$0.14$0.49$0.55$0.75$0.81$0.92$4.70$5.16$12.09$13.91
Latencythis run9.9 s2.1 s4.9 s16.5 s6 s2 s4.3 s3.1 s3.6 s4.4 s