Benchgen

BIG-Bench Hard — Results

RankModelScore
1deepseek-v4-1-flash86.1
2claude-3-5-sonnet0.931
3gemini-1-5-pro0.892
4claude-3-opus0.868
5gemini-1-5-flash0.855
6phi-3-5-moe-instruct0.791
7phi-4-mini0.704
8phi-3-5-mini-instruct0.69

BIG-Bench Hard — Phases

1 phaseActive

Phase 1·

BIG-Bench Hard

Aug 3, 2026
Current