MMMU authors

Massive Multi-discipline Multimodal Understanding

A broad multimodal reasoning benchmark spanning charts, diagrams, tables, and academic visual question answering.

In index
Multimodal
Best in multimodal
Source ↗

Models28

Top result87.2%Gemini 3 Pro

Top-3 spread1.2 pts

Year2024

Result and price

13 of 28 results

Results

28 results

#ModelResult

01Gemini 3 Pro87.2%

02GPT-5.286.7%

03Qwen3.6 Plus86.0%

04Qwen3.5 397B85.0%

05Kimi K2.584.3%

06Qwen3.5-122B-A10B83.9%

07Qwen3.6-27B82.9%

08Qwen3.5-27B82.3%

09Qwen3.6-35B-A3B81.7%

10Qwen3.5-35B-A3B81.4%

11Claude Opus 4.580.7%

12Claude 4 Opus Thinking76.5%

13Command A+75.1%

14Claude 3.7 Sonnet Thinking75.0%

15Claude 4 Opus73.7%

16Gemini 2.5 Flash-Lite72.9%

17Claude 3.7 Sonnet71.8%

18Nemotron 3 Nano Omni 30B A3B70.8%

19Gemini 2.0 Flash69.3%

20Grok 266.1%

21Gemini 2.0 Flash-Lite65.1%

22Mistral Small 3.164.0%

23Grok 2 mini63.2%

24Gemini 1.5 Flash58.3%

25LFM3-VL-2B55.7%

26LFM2.5-VL-3B48.4%

27ZAYA1-VL-8B46.0%

28LFM2.5-VL-450M32.7%