MMMU Benchmark — leaderboard
Massive Multi-discipline Multimodal Understanding benchmark. Tests academic visual reasoning across 30 subjects spanning science, engineering, art, and more.
Metric: Validation Score. Source: mmmu-benchmark.github.io. Status: saturated. 189 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Human Expert | 88.6 |
| 2 | GPT-5.1 | 85.4 |
| 3 | GPT-5 (Thinking) | 84.2 |
| 4 | O3 | 82.9 |
| 5 | O4 Mini | 81.6 |
| 6 | Claude Opus 4.5 | 80.7 |
| 7 | Gemini 2.5 Flash (Preview 05-20) | 79.7 |
| 8 | Qwen 3 VL 235B A22B | 78.7 |
| 9 | O1 | 78.2 |
| 10 | Grok 3 | 78 |
| 11 | Claude Sonnet 4.5 | 77.8 |
| 12 | Intern-S1 | 77.7 |
| 13 | Claude Opus 4.1 | 77.1 |
| 14 | Claude Opus 4 | 76.5 |
| 15 | Claude 3.7 Sonnet | 75 |
Interactive version: theaggregate.ai/benchmark?slug=mmmu-benchmark · How the rankings work · Data refreshed daily, snapshot 2026-07-22.