MMMU-Pro authors

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.

In index
Multimodal
Best in multimodal
Source ↗

Models53

Top result85.5%GPT-6.1 Terra

Top-3 spread1.6 pts

Year2024

Result and price

40 of 53 results

Results

53 results

#ModelResult

01GPT-6.1 Terra85.5%

02Kimi K3.1 Pro84.1%

03Gemini 3.1 Pro83.9%

04Gemini 3.5 Flash83.6%

05GPT-5.6 Sol83.0%

06Qwen3.8 Max82.3%

07Kimi K381.6%

07Inkling 2 Small81.6%

07Seed 2.1 Pro81.6%

10GPT-5.581.2%

10GPT-5.481.2%

12Gemini 3 Pro81.0%

13GPT-5.6 Terra80.7%

14Muse Spark80.4%

15Seed 2.1 Turbo80.1%

16GPT-5.279.5%

17Kimi K2.679.4%

18dots3-note Preview79.1%

19Qwen3.7 Plus79.0%

19Qwen3.5 397B79.0%

21Qwen3.6 Plus78.8%

22Hy478.6%

23Kimi K2.5 (Reasoning)78.5%

23Kimi K2.578.5%

25GPT-5.6 Luna78.4%

26MiniMax M378.1%

26Grok 4.378.1%

28MiMo-V2.577.9%

29Claude Opus 4.677.3%

30Qwen3.5-122B-A10B76.9%

30Gemma 4 31B76.9%

32GPT-5.4 mini76.6%

33Step 5 Preview76.0%

34Qwen3.6-27B75.8%

35Qwen3.6-35B-A3B75.3%

36Grok 4.2075.2%

37Qwen3.5-35B-A3B75.1%

38Qwen3.5-27B75.0%

39Inkling-Small74.0%

39Muse Glimmer 30B74.0%

41Gemma 4 26B A4B73.8%

42Inkling73.5%

43Interfaze Beta71.1%

44Claude Opus 4.570.6%

45Gemma 4 12B69.1%

46GPT-5.4 nano66.1%

47Command A+63.0%

48Gemma 4 E4B52.6%

49Qwen2.5-VL-32B49.5%

50Mistral Small 3.149.3%

51LFM3-VL-2B45.7%

52Gemma 4 E2B44.2%

53LFM2.5-VL-3B30.5%