Vellum - GPQA — leaderboard
Vellum's independently-run GPQA evaluation. Graduate-level science questions in physics, chemistry, and biology.
Metric: Accuracy (%). Source: www.vellum.ai. Status: saturated. 64 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 5 | 96.2 |
| 2 | Claude 3 Opus | 95.4 |
| 3 | Gemini 3.1 Pro (Preview) | 94.3 |
| 4 | Claude Opus 4.7 | 94.2 |
| 5 | Claude Fable 5 | 94.1 |
| 6 | Claude Mythos 5 | 94.1 |
| 7 | GPT-5.5 | 93.6 |
| 8 | Claude Opus 4.8 | 93.6 |
| 9 | MiniMax-M3 | 93 |
| 10 | GPT-5.2 | 92.4 |
| 11 | Gemini 3 Pro | 91.9 |
| 12 | Claude Opus 4.6 | 91.3 |
| 13 | GLM-5.2 | 91.2 |
| 14 | Kimi K2.6 | 90.5 |
| 15 | DeepSeek V4 Pro | 90.1 |
Interactive version: theaggregate.ai/benchmark?slug=vellum-gpqa · How the rankings work · Data refreshed daily, snapshot 2026-07-22.