Qwen3.7 Max

Qwen3

Ranks #36 of 43 on reading real contracts into structured billing data — -24 pts vs GPT-5.5.

45.3%

Accuracy

#36 of 43

19.6%

Hallucination

HIGH-confidence & wrong

$68.2

Cost / 1k contracts

$0.07 each · via OpenRouter

200.2s

Median latency

p90 349.9s

±13.3

Run-to-run σ

3 runs

0.66

Value

acc. pts per $/1k

Tokens per contract

Input22,889

Output7,785

Reasoning5,175

Reliability 94.4% · valid structured output across 18 calls.

Want the full picture? How we score →