Research2026-10-23
Ivo and micro1 tested seven AI models on 350 contract review tasks. All models scored below 60% on attorney-defined criteria. Grok 4.7 led with a 58.3% weighted pass rate. Opus 5.5 followed at 58.1% and Kimi K3 at 57.1%. Completion rates ranged from 80.8% for Opus to 100% for Kimi. Kimi cost 38% less per rollout than Grok. Kimi also finished 13% faster than Grok.
What changed
No prior benchmark existed for contract review performance across these models.
What it unlocks
Comparing seven leading AI models on 350 real-world contract review tasks with attorney-defined criteria.
- 58.3% weighted rubric pass rate for Grok 4.7
- 58.1% for Opus 5.5
- 57.1% for Kimi K3
- 80.8% to 100% completion rates
Sources