Claude Opus 5

NewClaude 5

Ranks #8 of 43 on reading real contracts into structured billing data — -0.9 pts vs GPT-5.5.

68.4%

Accuracy

#8 of 43

10.9%

Hallucination

HIGH-confidence & wrong

$346

Cost / 1k contracts

$0.35 each · via OpenRouter

67.1s

Median latency

p90 94.9s

±2.1

Run-to-run σ

3 runs

0.2

Value

acc. pts per $/1k

Tokens per contract

Input34,733

Output6,886

Reasoning426

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →