Harder, filtered variant of MMMU (Massive Multi-discipline Multimodal Understanding). Tests multimodal reasoning across 30 college-level subjects using image+text questions. The Pro subset removes questions that can be answered from text alone, isolating true multimodal ability.
This definition has not completed editorial and provenance review. Verify its metric, version, source, and evaluation configuration upstream before using it for a model-selection decision.
Filters out text-only-answerable questions for a cleaner multimodal signal; covers 30 subjects; broad model participation (70+ entries).
Some leaderboard entries allow tool use (Python execution) which inflates scores — this scraper prefers no-tools results to stay comparable across providers.
Higher is better. Percentage of multimodal questions answered correctly (overall score, no tools preferred).
Some leaderboard entries allow tool use (Python execution) which inflates scores — this scraper prefers no-tools results to stay comparable across providers.
| # | Model | Score |
|---|---|---|
| 1 | 81.2 | |
| 2 | 81.0 | |
| 3 | 80.5 | |
| 4 | 80.4 | |
| 5 | 79.5 | |
| 6 | 76.9 | |
| 7 | 76.4 | |
| 8 | 76.0 | |
| 9 | 74.5 | |
| 10 | 73.9 | |
| 11 | 73.9 | |
| 12 | 73.9 | |
| 13 | 73.8 | |
| 14 | 68.9 | |
| 15 | 68.9 | |
| 16 | 68.1 | |
| 17 | 68.0 | |
| 18 | 65.2 | |
| 19 | 62.7 | |
| 20 | 52.6 | |
| 21 | 51.9 | |
| 22 | 51.5 | |
| 23 | 46.2 | |
| 24 | 44.2 | |
| 25 | 43.5 | |
| 26 | 38.3 | |
| 27 | 37.6 | |
| 28 | 28.4 |