Kimi K2.6

NewKimi

Ranks #38 of 43 on reading real contracts into structured billing data — -26 pts vs GPT-5.5.

43.3%

Accuracy

#38 of 43

30.8%

Hallucination

HIGH-confidence & wrong

$90.2

Cost / 1k contracts

$0.09 each · via OpenRouter

437.7s

Median latency

p90 1362.9s

±13.9

Run-to-run σ

3 runs

0.48

Value

acc. pts per $/1k

Tokens per contract

Input21,028

Output17,560

Reasoning15,066

Reliability 94.4% · valid structured output across 18 calls.

Want the full picture? How we score →