everyais

gpqa-diamond

GPQA Diamond

Graduate-level science questions designed to resist simple retrieval.

Methodology

Scores are shown exactly as published by each source. Source boards are ranked independently and are never averaged.

Independent source results

Epoch AI Benchmarking Hub

Data source: Epoch AI Benchmarking HubLicense: CC-BY

Synced by EveryAI: This is the sync time, not the measurement date.

Epoch AI Benchmarking Hub leaderboard
#ModelScore
1Claude Opus 5everyais/claude-opus-592.9%
2MiniMax-M3everyais/minimax-m390.9%
GPQA Diamond | everyais