ModelVerdict

Christopher Scott

English (US)Text7 fields

Document en-p0003 · layout minutes · 7 fields · public part of the dataset · Same invoice in Czech → · Report an error in this example

Text (the model received exactly this text)

Meeting minutes – June 22, 2026

Attendees: Christopher Scott, Jeremy Bowen, Todd Davis, Lisa Richardson
For the vendor: Hunter, Clayton and Bond, EIN 91-9259754

1. The budget for next quarter was approved without changes.
2. Christopher Scott will compare the offers before the next meeting.
   Questions: call 437-555-0073.
3. Jeremy Bowen will confirm the delivery dates.
Minutes also sent to info@hunter-clayton.example.

Verdict

Cheapest with no errors: Gemma 4 31B at $0.08 per 1,000 documents.

  • Mistral Large 4 is also error-free, but 74.5× more expensive.

Prompt given to every model

Find all personal data of natural persons in the text below (for anonymisation) and return only
JSON with these keys, each a list of strings (an empty list when there is none):

person_names       full names of people, each person once, as written in the text
email_addresses    email addresses of people
phone_numbers      phone numbers of people, as written
postal_addresses   home or delivery addresses of people, as written (street, city, ZIP)
birth_dates        dates of birth (YYYY-MM-DD)
national_ids       national identification numbers (e.g. SSN, birth numbers)
bank_accounts      bank account numbers of people, as written

Rules:
- Only data of natural persons. Company names, company registration or tax numbers (EIN, IČO,
  VAT), generic company addresses such as info@ or support@, order and ticket numbers are not
  personal data: leave them out.
- Dates that are not dates of birth (meetings, deliveries, start dates) are not personal data.
- Do not invent or complete values; copy them from the text.

How we score

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

What each model extracted

FieldCorrect valueGemma 4 31B7 / 7 fieldsMistral Small 46 / 7 fieldsGPT-6 Luna7 / 7 fieldsClaude Haiku 5.57 / 7 fieldsDeepSeek V4.1 Flash7 / 7 fieldsQwen3.8 Flash7 / 7 fieldsGLM-5.3 Flash7 / 7 fieldsGemini 3.5 Flash Lite7 / 7 fieldsMistral Medium 3.57 / 7 fieldsGPT-6 Sol7 / 7 fieldsGemini 3.8 Flash7 / 7 fieldsClaude Sonnet 5.57 / 7 fieldsClaude Sonnet 57 / 7 fieldsMistral Large 47 / 7 fields
NamesChristopher Scott; Jeremy Bowen; Todd Davis; Lisa RichardsonChristopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct) Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson; Hunter; Clayton; Bondextra valueChristopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)Christopher Scott; Jeremy Bowen; Todd Davis; Lisa Richardson (correct)
Email addresses—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Phone numbers437-555-0073437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)437-555-0073 (correct)
Addresses—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Dates of birth—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
National IDs—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Bank accounts—— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)— (correct)
Tokensinput / output377 / 72354 / 84467 / 1991,104 / 108600 / 301677 / 607345 / 556365 / 105354 / 72467 / 135365 / 5571,104 / 1081,102 / 255553 / 2,727
including hidden reasoningbilled as output001410234506493007749801472,862
including cached inputinput billed at a discount24600036451200000000
Cost of this documentthis run$0.0001$0.0001$0.0001$0.0002$0.0003$0.0003$0.0003$0.0004$0.0011$0.0023$0.0024$0.0033$0.0048$0.0061
Per 1,000 documentsfor documents of this type$0.08$0.10$0.15$0.16$0.25$0.32$0.33$0.37$1.07$2.28$2.36$3.29$4.75$6.08
Latencythis run1.4 s0.9 s2.3 s1.6 s4.5 s13.7 s4.6 s0.7 s0.9 s3.8 s2.2 s3.1 s4.6 s70.2 s

Log in with email

We use your email only to log you in (and for the digest if you tick it). No spam.