Claude Fable 5

NewClaude 5

Ranks #4 of 43 on reading real contracts into structured billing data — +3.2 pts vs GPT-5.5.

72.5%

Accuracy

#4 of 43

7.8%

Hallucination

HIGH-confidence & wrong

$732

Cost / 1k contracts

$0.73 each · via OpenRouter

70.2s

Median latency

p90 120.2s

±4.8

Run-to-run σ

3 runs

0.1

Value

acc. pts per $/1k

Tokens per contract

Input34,733

Output7,702

Reasoning664

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →