Vals AI GPQA: leaderboard
Vals AI's own run of the 198-question GPQA Diamond set, four-option multiple choice questions in biology, physics and chemistry written and validated by subject experts. Models answer closed-book with their reasoning shown, and the selected letter is pulled out by pattern matching and compared with the key. The number is accuracy over the set, higher is better.
Metric: Accuracy (%). Source: www.vals.ai. Saturation forecast: Estimated already saturated. 138 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 95.45 | #58 (Gemini 3.1 Pro (Preview)) |
| 2 | GPT-5.6 Sol (Max) | 95.2 | #19 (GPT-5.6 Sol) |
| 3 | Grok 4.6 (High) | 94.7 | #26 (Grok 4.6) |
| 4 | Gemini 3.8 Flash (High) | 94.44 | #21 (Gemini 3.8 Flash) |
| 5 | Gemini 3.7 Flash (High) | 93.94 | #23 (Gemini 3.7 Flash) |
| 6 | Qwen 3.8 Max | 93.69 | #35 |
| 7 | Claude Opus 5 | 93.43 | #12 |
| 8 | Gemini 3.6 Flash (High) | 93.43 | #51 (Gemini 3.6 Flash) |
| 9 | Claude Fable 5.1 (Max) | 93.43 | #9 (Claude Fable 5.1) |
| 10 | GPT-5.5 (xHigh) | 93.18 | #31 (GPT-5.5) |
| 11 | Claude Fable 5 (Max) | 93.18 | #16 (Claude Fable 5) |
| 12 | Kimi K3 | 92.93 | #30 |
| 13 | Grok 4.5 (High) | 92.93 | #48 (Grok 4.5) |
| 14 | MiniMax-M3 | 92.68 | #136 |
| 15 | Gemini 3.5 Flash (High) | 92.68 | #60 (Gemini 3.5 Flash) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-gpqa · How It Works · Data refreshed daily, snapshot 2026-10-20.