Mistral Medium 3.5

NewMistral

Ranks #21 of 43 on reading real contracts into structured billing data — -6.4 pts vs GPT-5.5.

62.9%

Accuracy

#21 of 43

24.6%

Hallucination

HIGH-confidence & wrong

$63.4

Cost / 1k contracts

$0.06 each · via OpenRouter

62.4s

Median latency

p90 65.9s

±2.6

Run-to-run σ

3 runs

0.99

Value

acc. pts per $/1k

Tokens per contract

Input24,045

Output3,642

Reasoning0

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →