ModelVerdict

Omar Cabrera

English (US)Email9 fields

Document en-p0001 · layout plain · 9 fields · public part of the dataset · Same invoice in Czech → · Report an error in this example

Email text (the model got exactly this text)

From: Omar Cabrera <omar.cabrera@robinson.example>
To: support@smith-hughes.example
Date: Thu, Sep 18, 2025 at 5:20 PM
Subject: Billing error

Dear support team,

The amount on your latest invoice doesn't match our records, you charged us for one more item than we received. It's invoice 191813.

The disputed amount is 21 USD. Thank you for handling this. No rush on this.

Best regards,
Omar Cabrera
Robinson Inc

Verdict

Cheapest without an error: Gemma 4 31B at $0.11 per 1,000 documents.

  • Claude Sonnet 5 is also error-free but 25× more expensive.

Prompt given to every model

You triage the support inbox. Read the email below and return only JSON with these keys:

category        one of: billing (invoices, payments, charges), delivery (where is the order, shipping,
                address change), complaint (damaged, wrong or defective goods, warranty claims),
                technical (login, app or system errors, integrations), cancellation (cancel an order,
                contract or subscription), sales (quotes, prices, interest before buying)
priority        one of: urgent, high, normal, low (rules below)
sentiment       one of: positive, neutral, negative (the customer's tone)
customer_name   full name of the customer who wrote the request
order_number    the customer's order number, if mentioned
invoice_number  invoice number, if mentioned
amount          the amount of money the request is about (number)
currency        ISO 4217 code of that amount (USD, EUR, CZK, …)
deadline        the date by which the customer wants a resolution or answer (YYYY-MM-DD)

Priority rules:
- urgent: the customer cannot work right now, or the deadline is at most 1 day after the email date
- high: the deadline is at most 7 days after the email date, or the customer threatens to cancel
  or escalate (lawyer, ending the business)
- low: the customer says explicitly that it is not urgent
- normal: everything else

Rules:
- A thread: judge the newest message; order numbers, invoice numbers and amounts may come from the
  quoted earlier messages. A forwarded email: the customer is the original sender, not the
  colleague who forwarded it.
- Resolve relative deadlines ("by Friday", "by tomorrow", "by the end of the month") against the
  date of the message that states them.
- Support ticket numbers, tax IDs and phone numbers are not order or invoice numbers.
- If a value is not in the email, return null. Do not guess.
- 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 31B9 / 9 fieldsGPT-6 Luna8 / 9 fieldsMistral Small 47 / 9 fieldsQwen3.8 Flash9 / 9 fieldsDeepSeek V4.1 Flash9 / 9 fieldsGemini 3.5 Flash Lite9 / 9 fieldsMistral Medium 3.59 / 9 fieldsGPT-6 Sol9 / 9 fieldsGemini 3.8 Flash9 / 9 fieldsClaude Sonnet 59 / 9 fields
Categorybillingbilling (correct)billing (correct)billing (correct)billing (correct)billing (correct)billing (correct)billing (correct)billing (correct)billing (correct)billing (correct)
Prioritylowlow (correct)low (correct) normalwrong valuelow (correct)low (correct)low (correct)low (correct)low (correct)low (correct)low (correct)
Sentimentneutralneutral (correct) negativewrong value negativewrong valueneutral (correct)neutral (correct)neutral (correct)neutral (correct)neutral (correct)neutral (correct)neutral (correct)
Customer nameOmar CabreraOmar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)Omar Cabrera (correct)
Order number—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Invoice number191813191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)191813 (correct)
Amount21.0021.00 (correct)21 (correct)21 (correct)21 (correct)21 (correct)21.00 (correct)21 (correct)21.00 (correct)21 (correct)21.00 (correct)
CurrencyUSDUSD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)USD (correct)
Deadline—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Tokensinput / output588 / 68688 / 104558 / 67912 / 393828 / 218576 / 87558 / 67688 / 89576 / 526821 / 110
of which hidden reasoningbilled as output050030515400334410
of which from cacheinput billed at a discount40025615600000
Cost of this documentthis run$0.0001$0.0001$0.0001$0.0003$0.0003$0.0004$0.0013$0.0023$0.0024$0.0027
Per 1,000 documentsfor the same kind of document$0.11$0.12$0.12$0.29$0.29$0.39$1.34$2.27$2.40$2.74
Latencythis run2.7 s2.5 s1 s7.6 s3.8 s0.7 s0.7 s2.2 s2.2 s1.9 s