Epoch AI - GPQA Diamond: leaderboard
Epoch AI's own runs of GPQA Diamond, the 198 graduate-level biology, chemistry and physics multiple-choice questions, evaluated in Epoch's Inspect harness rather than taken from vendor reports. Each row is one Epoch run id, which keeps the reasoning effort or thinking budget used; score is the best accuracy across Epoch's scorers in percent, higher is better, random guessing 25. Evaluator-distinct from the vendor-reported GPQA Diamond boards.
Metric: Accuracy (%). Source: epoch.ai. Saturation forecast: Estimated already saturated. 333 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon (High) | 96.4 | #9 (Gemini 4 Argon) |
| 2 | GPT-6 Astra (Max) | 95.77 | #6 (GPT-6 Astra) |
| 3 | Claude Sonnet 5.5 (Max) | 95.58 | #14 (Claude Sonnet 5.5) |
| 4 | Gemini 3.8 Flash (High) | 95.39 | #30 (Gemini 3.8 Flash) |
| 5 | GPT-6.1 Sol (Max) | 95.39 | #7 (GPT-6.1 Sol) |
| 6 | Gemini 3.7 Flash (High) | 94.82 | #24 (Gemini 3.7 Flash) |
| 7 | Gemini 3.1 Pro (Preview) (High) | 94.44 | #64 (Gemini 3.1 Pro (Preview)) |
| 8 | GPT-6 Sol (Max) | 94.32 | #19 (GPT-6 Sol) |
| 9 | GPT-6.1 Terra (Max) | 94.32 | #15 (GPT-6.1 Terra) |
| 10 | Gemini 3.6 Flash (High) | 94.13 | #57 (Gemini 3.6 Flash) |
| 11 | Gemini 3.1 Pro (Preview) | 94.1 | #64 |
| 12 | GPT-5.5 (xHigh) | 94 | #33 (GPT-5.5) |
| 13 | Grok 4.6 (High) | 94 | #29 (Grok 4.6) |
| 14 | GPT-5.5 Pro (xHigh) | 93.92 | #23 (GPT-5.5 Pro) |
| 15 | Claude Opus 5 (Max) | 93.88 | #16 (Claude Opus 5) |
No result here: #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=epoch-ai-gpqa-diamond · How It Works · Data refreshed daily, snapshot 2026-11-05.