What it measures
- Multimodal academic QA
- Image-text reasoning
- Expert subject knowledge
MMMU-Pro
A harder multimodal academic benchmark derived from MMMU-style tasks.
How to read the score
% accuracy; higher is better.
Sourced Ranking
| Rank | Model | Provider | Score | Source | Source Type | Verified |
|---|---|---|---|---|---|---|
| 1 | OpenAI GPT-5.4 | OpenAI | 81.2 | Introducing GPT-5.4 Accessed 2026-06-11 | official | Yes |
| 2 | Gemini 3 Pro | 81 | Gemini 3.1 Pro model page Accessed 2026-06-11 | official | Yes | |
| 3 | Gemini 3.1 Pro | 80.5 | Gemini 3.1 Pro model page Accessed 2026-06-11 | official | Yes | |
| 4 | OpenAI GPT-5.2 Thinking | OpenAI | 79.5 | Introducing GPT-5.2 Accessed 2026-06-11 | official | Yes |
Limitations
- New high scores may appear on model release pages before centralized boards update.
- The arXiv paper is static; leaderboard and Hugging Face sources are better.
- MMMU official page and HF dataset are selected entry points.
leaderboard
MMMU official page, Hugging Face dataset, and third-party score boards
The arXiv paper is static; leaderboard and Hugging Face sources are better. MMMU official page and HF dataset are selected entry points.
paper
MMMU-Pro paper or specification
Use for benchmark definition, not necessarily for latest model scores.
official
MMMU-Pro dataset or code
Use for dataset, code, or implementation details; score freshness depends on the benchmark.