// benchmark explainer
MMMU / MMMU-Pro, explained
By the TrueIQ editorial team at Ahzy Capital · Last reviewed Oct 15, 2026 · How we work
MMMU tests whether a model can answer college level questions that combine text with images, such as charts, diagrams, chemical structures and sheet music. MMMU-Pro is a harder, cleaned up version.
| Made by | MMMU team (2023); MMMU-Pro (2024) |
|---|---|
| Size | About 11,500 questions across 30 subjects in 6 disciplines |
| Format | Mostly multiple choice |
| Paper | arXiv 2311.16502 |
What it measures
It checks real visual understanding combined with subject knowledge, not just describing a picture. MMMU-Pro removes questions that can be answered without the image and adds more answer options.
Where it falls short
- Multiple choice lets models use elimination.
- Image quality and resolution settings affect scores.
- Public questions can leak into training data.
How to read a score
Say which version you mean (MMMU or MMMU-Pro). The scores aren't comparable.
Where to see current scores
We don't republish scores. Leaderboards change weekly, and numbers depend on how the test was run. Check these instead:
- Artificial Analysis: MMMU / MMMU-Pro results (independently run)
- MMMU official site
- Paper