Multimodal

MMMU

Tests the AI on understanding images and text together across many college subjects. Higher is better.

Rankings

Higher is better

MMMU — frequently asked questions

What is MMMU?
Tests the AI on understanding images and text together across many college subjects. Higher is better.
Which AI model scores highest on MMMU?
Qwen3.5 by Qwen holds the best MMMU result among tracked models, at 85% (released Feb 16 2026). Higher scores are better on this benchmark.
What are the top 5 models on MMMU?
1. Qwen3.5 (Qwen) — 85%; 2. Qwen3.6 (Qwen) — 81.7%; 3. Gemini 3.0 Pro (Google) — 81%; 4. Muse Spark (Meta) — 80.4%; 5. GPT-5.1 (OpenAI) — 76%.
How many models have a published MMMU score?
7 tracked models have a published MMMU score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.
What is the best open model on MMMU?
Qwen3.5 by Qwen is the highest-ranked model with downloadable weights on MMMU, scoring 85% at rank 1 overall.
← All benchmarks