MMMU-Pro
multimodal
+
+
+
+
About
MMMU-Pro is an enhanced version of the Massive Multi-discipline Multimodal Understanding benchmark featuring more challenging expert-level questions and advanced reasoning requirements. This professional-grade variant pushes the boundaries of multimodal evaluation with increased complexity in visual reasoning, domain-specific knowledge application, and multi-step problem solving across academic disciplines.
+
+
+
+
Evaluation Stats
Total Models12
Organizations4
Verified Results0
Self-Reported12
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
12 models
Top Score
78.4%
Average Score
51.1%
High Performers (80%+)
0Top Organizations
#1OpenAI
3 models
71.6%
#2Meta
3 models
45.9%
#3Alibaba Cloud / Qwen Team
5 models
44.3%
#4Microsoft
1 model
38.5%
+
+
+
+
Leaderboard
12 models ranked by performance on MMMU-Pro
| License | Links | ||||
|---|---|---|---|---|---|
| Aug 7, 2025 | Proprietary | 78.4% | |||
| Apr 16, 2025 | Proprietary | 76.4% | |||
| Aug 6, 2024 | Proprietary | 59.9% | |||
| Apr 5, 2025 | Llama 4 Community License Agreement | 59.6% | |||
| Jan 26, 2025 | tongyi-qianwen | 51.1% | |||
| Feb 28, 2025 | Apache 2.0 | 49.5% | |||
| Aug 29, 2024 | tongyi-qianwen | 46.2% | |||
| Sep 25, 2024 | Llama 3.2 | 45.2% | |||
| Feb 1, 2025 | MIT | 38.5% | |||
| Jan 26, 2025 | Apache 2.0 | 38.3% |
Showing 1 to 10 of 12 models