Skip to content

Massive Multi-discipline Multimodal Understanding Pro

A harder multimodal benchmark for frontier models that combines text with images, diagrams, charts, and academic visual reasoning tasks.

48 models

Higher is better.

  1. 1Gemini 3.1 ProGoogle83.9
  2. 2Gemini 3.5 FlashGoogle83.6
  3. 3GPT-5.6 SolOpenAI83.0
  4. 4Qwen3.8 MaxAlibaba82.3
  5. 5Kimi K3Moonshot AI81.6
  6. 6Seed 2.1 ProByteDance81.6
  7. 7GPT-5.5OpenAI81.2
  8. 8GPT-5.4OpenAI81.2
  9. 9Gemini 3 ProGoogle81.0
  10. 10GPT-5.6 TerraOpenAI80.7
  11. 11Muse SparkMeta80.4
  12. 12GPT-5.2OpenAI79.5
  13. 13Kimi K2.6Moonshot AI79.4
  14. 14dots3-note PreviewDots Studio79.1
  15. 15Qwen3.5 397BAlibaba79.0
  16. 16Qwen3.7 PlusAlibaba79.0
  17. 17Qwen3.6 PlusAlibaba78.8
  18. 18Kimi K2.5 (Reasoning)Moonshot AI78.5
  19. 19Kimi K2.5Moonshot AI78.5
  20. 20GPT-5.6 LunaOpenAI78.4
  21. 21MiniMax M3MiniMax78.1
  22. 22Grok 4.3xAI78.1
  23. 23Pareto 26.9Unbiased78.0
  24. 24MiMo-V2.5Xiaomi77.9
  25. 25Claude Opus 4.6Anthropic77.3
  26. 26Qwen3.5-122B-A10BAlibaba76.9
  27. 27Gemma 4 31BGoogle76.9
  28. 28GPT-5.4 miniOpenAI76.6
  29. 29Step 5 PreviewStepFun76.0
  30. 30Qwen3.6-27BAlibaba75.8
  31. 31Qwen3.6-35B-A3BAlibaba75.3
  32. 32Grok 4.20xAI75.2
  33. 33Qwen3.5-35B-A3BAlibaba75.1
  34. 34Qwen3.5-27BAlibaba75.0
  35. 35Inkling-SmallThinking Machines Lab74.0
  36. 36Muse Glimmer 30BMeta74.0
  37. 37Gemma 4 26B A4BGoogle73.8
  38. 38InklingThinking Machines Lab73.5
  39. 39Interfaze BetaInterfaze71.1
  40. 40Claude Opus 4.5Anthropic70.6
Higher is better. Showing the top 40 of 48. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Image + text question answering
Tasks
Multimodal academic reasoning
Authors
MMMU-Pro authors (2024)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.

All results

1Gemini 3.1 Pro Google83.9
2Gemini 3.5 Flash Google83.6
3GPT-5.6 Sol OpenAI83.0
4Qwen3.8 Max Alibaba82.3
5Kimi K3 Moonshot AI81.6
6Seed 2.1 Pro ByteDance81.6
7GPT-5.5 OpenAI81.2
8GPT-5.4 OpenAI81.2
9Gemini 3 Pro Google81.0
10GPT-5.6 Terra OpenAI80.7
11Muse Spark Meta80.4
12GPT-5.2 OpenAI79.5
13Kimi K2.6 Moonshot AI79.4
14dots3-note Preview Dots Studio79.1
15Qwen3.5 397B Alibaba79.0
16Qwen3.7 Plus Alibaba79.0
17Qwen3.6 Plus Alibaba78.8
18Kimi K2.5 (Reasoning) Moonshot AI78.5
19Kimi K2.5 Moonshot AI78.5
20GPT-5.6 Luna OpenAI78.4
21MiniMax M3 MiniMax78.1
22Grok 4.3 xAI78.1
23Pareto 26.9 Unbiased78.0
24MiMo-V2.5 Xiaomi77.9
25Claude Opus 4.6 Anthropic77.3
26Qwen3.5-122B-A10B Alibaba76.9
27Gemma 4 31B Google76.9
28GPT-5.4 mini OpenAI76.6
29Step 5 Preview StepFun76.0
30Qwen3.6-27B Alibaba75.8
31Qwen3.6-35B-A3B Alibaba75.3
32Grok 4.20 xAI75.2
33Qwen3.5-35B-A3B Alibaba75.1
34Qwen3.5-27B Alibaba75.0
35Inkling-Small Thinking Machines Lab74.0
36Muse Glimmer 30B Meta74.0
37Gemma 4 26B A4B Google73.8
38Inkling Thinking Machines Lab73.5
39Interfaze Beta Interfaze71.1
40Claude Opus 4.5 Anthropic70.6
41Gemma 4 12B Google69.1
42GPT-5.4 nano OpenAI66.1
43Command A+ Cohere63.0
44Gemma 4 E4B Google52.6
45Qwen2.5-VL-32B Alibaba49.5
46Mistral Small 3.1 Mistral49.3
47Gemma 4 E2B Google44.2
48LFM2.5-VL-3B Liquid AI30.5