Benchmark

GPQA Diamond (Epoch)

Epoch AI · source page ↗ · last checked Aug 24, 2026, 12:02 PM

Current leaderboard

captured Aug 24, 2026, 12:02 PM
ModelGPQA Diamond accuracy (0–1)
gemini-3.7-flash_high0.948
gpt-5.4-pro-2026-03-05_xhigh0.946
gemini-3.1-pro-preview_high0.944
gemini-3.6-flash_high0.941
gemini-3.1-pro-preview0.941
grok-4.6_high0.94
gpt-5.5-pre-release_xhigh0.94
claude-opus-5_max0.939
gpt-5.5-pro-pre-release_xhigh0.939
gpt-5.6-sol_max0.935
grok-4.5_high0.934
gpt-5.6-terra_max0.933
gpt-5.4-2026-03-05_xhigh0.933
grok-4.6_xhigh0.932
kimi-k3_max0.931
claude-opus-50.929
gemini-3.5-flash_high0.928
qwen3.8-max_xhigh0.927
gemini-3-pro-preview0.926
kimi-k3_high0.919
glm-5.2_max0.919
deepseek-v4-pro-0813_max0.917
gpt-5.6-luna_max0.916
gpt-5.2-2025-12-11_xhigh0.914
deepseek-v4-flash-0731_max0.91

Change history