Vals AI MMLU-Pro: leaderboard
Metric: Accuracy (%). Source: www.vals.ai. 138 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 92.38 |
| 2 | Claude Opus 5 (Max) | 91.59 |
| 3 | Claude Fable 5 (Max) | 91.5 |
| 4 | Gemini 3.1 Pro (Preview) (High) | 90.99 |
| 5 | Gemini 3.8 Flash (High) | 90.22 |
| 6 | Gemini 3.7 Flash (High) | 90.12 |
| 7 | Gemini 3 Pro (Preview) | 90.1 |
| 8 | Claude Opus 4.7 (Max) | 89.87 |
| 9 | Claude Opus 4.8 (Max) | 89.58 |
| 10 | Gemini 3.5 Flash (High) | 89.52 |
| 11 | Grok 4.6 (High) | 89.4 |
| 12 | Qwen 3.7 Max | 89.31 |
| 13 | Gemini 3.6 Flash (High) | 89.28 |
| 14 | Grok 4.5 (High) | 89.22 |
| 15 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 89.11 |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-mmlu-pro · How It Works · Data refreshed daily, snapshot 2026-09-05.