Mistral Medium 3.5

NewMistral

Ranks #13 of 35 on reading real contracts into structured billing data, -4.9 pts vs GPT-5.5.

71.1%

Accuracy

#13 of 35

10.8%

Hallucination

HIGH-confidence & wrong

$66.1

Cost / 1k contracts

$0.07 each · via OpenRouter

26.6s

Median latency

p90 45.7s

±0

Run-to-run σ

1 runs

1.08

Value

acc. pts per $/1k

Tokens per contract

Input24,977

Output3,816

Reasoning0

Reliability 100% · valid structured output across 30 calls.

Want the full picture? How we score