unbiased.ai
Contest #003Ran 2026-07-04
Contest #003 · The Extraction Drag Race · Four models

Four AIs read the same 12 messy docs.
The priciest was the least accurate.

Same 177 fields — smudged receipts, two-column resumes, a handwritten contract override. One shot each. Accuracy landed a coin flip apart: 98.9% vs 93.2%. The bills didn't. Guess which one ran up the biggest bill, then reveal the receipts.

A
97.2%
accuracy
$•.••••
B
98.9%
accuracy
$•.••••
C
93.2%
accuracy
$•.••••
D
98.9%
accuracy
$•.••••
All four cleared 93%+ on the same 177-field rubric. Guess the priciest — then reveal.
$0.0335
The cheapest of the four (97.2% accurate)
$0.1499
The priciest — and the least accurate (93.2%)
4.5×
Price gap across a 5.7-point accuracy spread
4 / 4
Handwriting traps every model caught cleanly

The bill has nothing to do with the accuracy.

Four models, 177 fields each. Every one landed between 93.2% and 98.9% — a coin flip apart. Yet the priciest cost 4.5× the cheapest, and scored dead last.

Most extraction pipelines are hard-wired to one expensive model, on faith that the price buys accuracy. On these 12 documents it didn't. The premium bought nothing you can measure — it was pure overpay.

That's what Pareto fixes. It routes every document to the cheapest model that clears your accuracy bar, on one endpoint you never have to think about. You land on the low number without giving up a single field.
Run 1M documents a month
Hard-coded to the priciest (93.2%)$12,492
Routed to the cheapest that clears the bar$2,792
You'd save — for equal-or-better accuracy
$9,700/mo
$116,400 a year
Per-document cost ($0.1499 vs $0.0335 ÷ 12 docs) × volume. Illustrative — one workload; your real mix varies. Measuring that mix is exactly what the free audit does.

Accuracy, to the tenth of a point

Share of 177 leaf fields extracted correctly across all 12 documents. Blind labels A–D.

A
97.2%172 / 177 fields
B
98.9%175 / 177 fields
C
93.2%165 / 177 fields
D
98.9%175 / 177 fields
Top to bottom: 5.7 points. The amber bar (C) is the most expensive model in the field — and the least accurate. Bar length is proportional; nothing is exaggerated.
The credibility star · Document 9

The handwritten override that separates reading from parsing

A logistics services agreement printed a 30-day termination notice. Someone crossed it out by hand and wrote in 60, then initialed it. A model that only reads the typed text gets this wrong. The correct answer is 60 — and every model had to see the ink to know it.

What was printed
§7.2  Either party may terminate this
Agreement upon thirty (30) days'
written notice to the other party.
What was handwritten in the margin
changed to 60 days
   — per rider, initialed J.M.

Truth = 60 days
What each blind model extracted for termination_notice_days
A
60
✓ caught
B
60
✓ caught
C
60
✓ caught
D
60
✓ caught
All four caught it — including the inverse trap in Document 10, where a handwritten note confirmed "90 days is fine — NO change" and the value had to stay 90. Reading the ink is table stakes now. The differentiator wasn't whether they could read — it was what they charged to.

The verdict

Every model read the documents. Smudged OCR, GBP vs USD, balance-forward and credit-memo totals, two-column resumes, both handwriting traps — all four landed within 5.7 points on the same 177-field rubric. On accuracy it's effectively a coin flip.

So the thing that most separates them is the bill, and the bill ranges 4.5×. The most expensive model in the field scored last. You paid a premium for a worse answer.

Pick by brand and you're gambling with your budget. Pick the cheapest that still clears the bar — which is all routing is — and you keep the accuracy while cutting the bill to a fraction. "We'll send every doc to whatever's cheapest that works" is a sentence no single lab can say. We can.

The receipts

Cost to extract all 12 documents once. Accuracy in parentheses. Blind labels — we don't map model to letter.

ModelCostvs cheapest
A (97.2%)CHEAPEST$0.03351.0×
B (98.9%)$0.04251.3×
D (98.9%)$0.08912.7×
C (93.2%)PRICIEST$0.14994.5×

Four leading models ran this — Claude Fable 5, Claude Opus 4.8, GPT-5.5, and one more. We're not telling you which is which, because on accuracy you couldn't tell either. That's the point.

One shot each

Identical prompt and identical 12 documents, no retries, no cherry-picking. Each result is exactly what the model returned.

Graded against ground truth

177 leaf fields with a fixed rubric: case-insensitive strings, numbers to 2dp, dates ISO. Accuracy = fields correct ÷ 177.

Real bills, one yardstick

Two models ran as Claude agent runs (output chars ÷ 4 × published rate); GPT-5.5 and one more billed on real usage tokens. Same yardstick for all four.

Labels withheld on purpose

We don't map model to letter. If price bought accuracy you'd spot the expensive one on the bars. You can't — because it didn't.

Ran 2026-07-04 · same prompt, one shot each · Two models ran as Claude agent runs via the Anthropic API; GPT-5.5 and one more ran through our at-cost gateway · Labels withheld on purpose.

Send us your AI bill.

Your extraction pipeline has a spread like this too. We route every document to the cheapest model that clears your accuracy bar — and prove it nightly. What you spent, what you'd have spent, receipts attached.