Ministral 14B

Mistral

Ranks #29 of 43 on reading real contracts into structured billing data — -12.8 pts vs GPT-5.5.

56.5%

Accuracy

#29 of 43

9.8%

Hallucination

HIGH-confidence & wrong

$5.50

Cost / 1k contracts

$0.01 each · via OpenRouter

60.9s

Median latency

p90 68.2s

±6.3

Run-to-run σ

3 runs

10.26

Value

acc. pts per $/1k

Tokens per contract

Input24,049

Output3,472

Reasoning0

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →