Grok 4.3

Grok 4

Ranks #25 of 43 on reading real contracts into structured billing data — -9.5 pts vs GPT-5.5.

59.8%

Accuracy

#25 of 43

24.7%

Hallucination

HIGH-confidence & wrong

$33.7

Cost / 1k contracts

$0.03 each · via OpenRouter

24.6s

Median latency

p90 34.8s

±9.1

Run-to-run σ

3 runs

1.78

Value

acc. pts per $/1k

Tokens per contract

Input20,784

Output3,083

Reasoning982

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →