Omar Cabrera
Texto del correo (el modelo recibió exactamente este texto)
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
Veredicto
El más barato sin errores: Gemma 4 31B, a 0,10 € por cada 1000 documentos.
- Claude Sonnet 5 tampoco comete errores, pero es 25× más caro.
Prompt enviado a todos los modelos
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.
Cómo puntuamos
Coincidencia exacta por campo tras normalizar fechas, importes y espacios. La falta de diacríticos cuenta como error. El coste es el gasto real en tokens de esta ejecución, convertido al tipo de cambio indicado en el pie de página.
Lo que extrajo cada modelo
| Campo | Valor correcto | Gemma 4 31B9 / 9 campos | GPT-6 Luna8 / 9 campos | Mistral Small 47 / 9 campos | Qwen3.8 Flash9 / 9 campos | DeepSeek V4.1 Flash9 / 9 campos | Gemini 3.5 Flash Lite9 / 9 campos | Mistral Medium 3.59 / 9 campos | GPT-6 Sol9 / 9 campos | Gemini 3.8 Flash9 / 9 campos | Claude Sonnet 59 / 9 campos |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Categoría | billing | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) | billing (correcto) |
| Prioridad | low | low (correcto) | low (correcto) | normalvalor incorrecto | low (correcto) | low (correcto) | low (correcto) | low (correcto) | low (correcto) | low (correcto) | low (correcto) |
| Sentimiento | neutral | neutral (correcto) | negativevalor incorrecto | negativevalor incorrecto | neutral (correcto) | neutral (correcto) | neutral (correcto) | neutral (correcto) | neutral (correcto) | neutral (correcto) | neutral (correcto) |
| Nombre del cliente | Omar Cabrera | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) | Omar Cabrera (correcto) |
| Número de pedido | — | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) |
| Número de factura | 191813 | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) | 191813 (correcto) |
| Importe | 21.00 | 21.00 (correcto) | 21 (correcto) | 21 (correcto) | 21 (correcto) | 21 (correcto) | 21.00 (correcto) | 21 (correcto) | 21.00 (correcto) | 21 (correcto) | 21.00 (correcto) |
| Moneda | USD | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) | USD (correcto) |
| Plazo | — | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) | — (correcto) |
| Tokens | entrada / salida | 588 / 68 | 688 / 104 | 558 / 67 | 912 / 393 | 828 / 218 | 576 / 87 | 558 / 67 | 688 / 89 | 576 / 526 | 821 / 110 |
| de ellos, razonamiento oculto | se factura como salida | 0 | 50 | 0 | 305 | 154 | 0 | 0 | 33 | 441 | 0 |
| de ellos, desde caché | entrada con descuento | 4 | 0 | 0 | 256 | 156 | 0 | 0 | 0 | 0 | 0 |
| Coste de este documento | esta ejecución | 0,0001 € | 0,0001 € | 0,0001 € | 0,0003 € | 0,0003 € | 0,0004 € | 0,0012 € | 0,0021 € | 0,0022 € | 0,0025 € |
| Por cada 1000 documentos | para el mismo tipo de documento | 0,10 € | 0,11 € | 0,11 € | 0,26 € | 0,27 € | 0,36 € | 1,23 € | 2,08 € | 2,21 € | 2,52 € |
| Latencia | esta ejecución | 2,7 s | 2,5 s | 1 s | 7,6 s | 3,8 s | 0,7 s | 0,7 s | 2,2 s | 2,2 s | 1,9 s |