Qwen3.8 Max

NewQwen3

Ranks #31 of 43 on reading real contracts into structured billing data — -15.5 pts vs GPT-5.5.

53.8%

Accuracy

#31 of 43

26%

Hallucination

HIGH-confidence & wrong

$199

Cost / 1k contracts

$0.20 each · via OpenRouter

503.4s

Median latency

p90 600.4s

±7.3

Run-to-run σ

3 runs

0.27

Value

acc. pts per $/1k

Tokens per contract

Input22,930

Output25,488

Reasoning21,203

Reliability 94.4% · valid structured output across 18 calls.

Want the full picture? How we score →