MMMU measures multimodal academic reasoning across college-level questions with images, diagrams, and charts. It is a broad vision-language benchmark for cross-discipline capability checks.
Broad coverage across disciplines and subjects gives a strong general multimodal signal; questions require combining visual perception with domain knowledge.
Many tasks emphasize perception plus recall more than deep multi-step visual planning; OCR/text cues can let some models score without full visual grounding.
Higher is better. Accuracy percentage.
Max score: 100
Many tasks emphasize perception plus recall more than deep multi-step visual planning.
| # | Model | Score |
|---|---|---|
| 1 | 87.8 | |
| 2 | 87.6 | |
| 3 | 87.6 | |
| 4 | 87.5 | |
| 5 | 87.5 | |
| 6 | 87.5 | |
| 7 | 87.4 | |
| 8 | 87.3 | |
| 9 | 87.2 | |
| 10 | 86.7 | |
| 11 | 86.7 | |
| 12 | 86.7 | |
| 13 | 86.7 | |
| 14 | 86.3 | |
| 15 | 86.2 | |
| 16 | 86.2 | |
| 17 | 86.1 | |
| 18 | 85.7 | |
| 19 | 85.6 | |
| 20 | 85.6 | |
| 21 | 85.6 | |
| 22 | 85.5 | |
| 23 | 85.4 | |
| 24 | 85.4 | |
| 25 | 85.4 | |
| 26 | 85.3 | |
| 27 | 85.2 | |
| 28 | 85.1 | |
| 29 | 85.0 | |
| 30 | 85.0 | |
| 31 | 84.8 | |
| 32 | 84.8 | |
| 33 | 84.5 | |
| 34 | 84.5 | |
| 35 | 84.3 | |
| 36 | 84.2 | |
| 37 | 84.2 | |
| 38 | 84.2 | |
| 39 | 84.2 | |
| 40 | 84.2 | |
| 41 | 84.2 | |
| 42 | 84.2 | |
| 43 | 84.0 | |
| 44 | 84.0 | |
| 45 | 84.0 | |
| 46 | 84.0 | |
| 47 | 84.0 | |
| 48 | 84.0 | |
| 49 | 83.9 | |
| 50 | 83.8 |