Vals AI

MMMU Pro

Multimodal Multi-task Benchmark

In index
Multimodal
Best in multimodal
Source ↗

Models92

Top result90.6%Claude Fable 5.1

Top-3 spread1.3 pts

Updated1 Sep 2026

Result and cost

92 of 92 results

Results

92 results

#ModelResultCostTime

01Claude Fable 5.190.6% ±0.7$0.16637 s

02Claude Opus 589.9% ±0.7$0.06930 s

03Claude Fable 589.3% ±0.7$0.16161 s

04Gemini 3.8 Flash89.1% ±0.8$0.004512 s

05Gemini 3.7 Flash89.0% ±0.8$0.0146.3 s

06GPT-5.6 Sol88.8% ±0.8$0.05626 s

07Gemini 3.6 Flash88.4% ±0.8$0.01911 s

08GPT-5.588.3% ±0.8$0.04354 s

08Gemini 3.5 Flash88.3% ±0.8$0.02512 s

10Gemini 3.1 Pro Preview88.2% ±0.8$0.06477 s

10Kimi K388.2% ±0.8$0.0592 min

12Qwen3.8 Max88.0% ±0.8$0.01644 s

13Gemini 3 Flash Preview87.6% ±0.8$0.01428 s

14GPT-5.487.5% ±0.8$0.04766 s

14Gemini 3 Pro87.5% ±0.8$0.0148 s

16Muse Spark87.4% ±0.8$0.001748 s

17GPT-5.286.7% ±0.8$0.0612 min

18Claude Opus 4.886.6% ±0.8$0.12149 s

18Muse Spark 1.186.6% ±0.8$0.01432 s

20GPT-5.6 Terra86.5% ±0.8$0.01613 s

21Kimi K2.686.3% ±0.8$0.0354 min

22Muse Spark 1.286.1% ±0.8$0.01874 s

23GLM-5.3-Flash86.0% ±0.8$0.0009628 s

24Claude Opus 4.785.5% ±0.8$0.1771 s

25GPT-5.6 Luna85.0% ±0.9$0.006435 s

26Kimi K2.5 Thinking84.3% ±0.9$0.022 min

27Qwen3.6 Plus84.2% ±0.9$0.0172 min

28Qwen3.8-27B83.9% ±0.9$0.01343 s

28Claude Opus 4.6 (Adaptive)83.9% ±0.9$0.1162 min

30Claude Sonnet 4.683.6% ±0.9$0.1442 min

30Gemini 3.5 Flash-Lite83.6% ±0.9$0.00667.8 s

32Grok 4.20 (Reasoning)83.5% ±0.9$0.02744 s

33GPT-5.183.2% ±0.9$0.002875 s

34Grok 4.383.1% ±0.9$0.0223 min

35Claude Sonnet 583.0% ±0.9$0.03319 s

36Claude Opus 4.5 Thinking82.9% ±0.9$0.12775 s

37Gemini 3.1 Flash Lite Preview82.5% ±0.9$0.003212 s

38Qwen3.5 Flash81.9% ±0.9$0.002646 s

39GPT-581.5% ±0.9$0.03372 s

40Gemini 2.5 Pro Exp 03.2581.3% ±0.9$0.007824 s

41MiniMax M381.2% ±0.9$0.009468 s

42Claude Opus 4.581.1% ±0.9$0.02220 s

43Gemini 2.5 Flash Preview Thinking80.8% ±0.9$0.001928 s

44o380.4% ±1.0$0.01539 s

45MiMo-V2.580.0% ±1.0$0.001442 s

46o4-mini79.7% ±1.0$0.01120 s

47Gemini 2.5 Flash (Preview)79.5% ±1.0$0.001926 s

48Claude Sonnet 4.5 Thinking79.3% ±1.0$0.05856 s

49GPT-5.4 mini79.2% ±1.0$0.007151 s

50GPT-5 mini78.9% ±1.0$0.007227 s

51Inkling78.5% ±1.0$0.0312 min

52Claude Opus 4.1 Thinking77.5% ±1.0$0.1692.4 s

53o177.4% ±1.0$0.14626 s

54Grok 476.3% ±1.0$0.0982 min

55Gemini 2.5 Flash Lite Preview Thinking75.4% ±1.0$0.0003626 s

56Claude 3.7 Sonnet Thinking75.1% ±1.0$0.05170 s

57Claude Sonnet 4 Thinking74.9% ±1.0$0.05149 s

58Claude Opus 4.173.7% ±1.1$0.04446 s

59GPT-5.4 nano73.6% ±1.1$0.0008422 s

60Claude Opus 473.3% ±1.1$0.06214 s

61Grok 4 Fast (Reasoning)72.8% ±1.1$0.002919 s

62Grok 4.1 Fast (Reasoning)72.7% ±1.1$0.000562 min

63Gemini 2.5 Flash Lite Preview72.5% ±1.1$0.000920 s

64Claude Sonnet 472.4% ±1.1$0.01311 s

64GPT-4.172.4% ±1.1$0.00812 s

66Gemini 2.5 Flash Preview 04.17 Thinking71.9% ±1.1$0.00268 min

67Llama4 Maverick Instruct Basic71.7% ±1.1$0.00117.9 s

68Claude 3.7 Sonnet71.5% ±1.1$0.0128.3 s

69GPT-5 nano70.9% ±1.1$0.002631 s

70GPT-4.1 mini70.5% ±1.1$0.00188.2 s

71Gemini 2.0 Flash 00169.8% ±1.1$0.0003745 s

72Claude 3.5 Sonnet68.8% ±1.1$0.0146 s

73Gemini 2.5 Flash Preview 04.1768.0% ±1.1$0.02537 s

74Mistral Large 366.2% ±1.1$0.001514 s

75Gemini 1.5 Pro 00265.5% ±1.1$0.001844 s

76Magistral Small65.2% ±1.1$0.004419 s

77Magistral Medium64.6% ±1.1$0.02142 s

78Grok 4.1 Fast Non Reasoning63.7% ±1.2$0.0006351 s

79Grok 4 Fast Non Reasoning63.4% ±1.2$0.00167.2 s

80Mistral Medium63.0% ±1.2$0.001613 s

81GPT-4o62.2% ±1.2$0.006846 s

82Grok 4.561.8% ±1.2$0.0312 min

83Mistral Small 3.160.1% ±1.2$0.0002713 s

84Llama 4 Scout 17B 16E Instruct58.8% ±1.2$0.0008411 s

85Grok 2 Vision57.3% ±1.2$0.005418 s

86Gemini 1.5 Flash 00257.2% ±1.2$0.0001141 s

87GPT-4o mini56.6% ±1.2$0.0003745 s

88GPT-4.1 nano55.1% ±1.2$0.000577.3 s

89Llama 3.2 90B Vision Instruct Turbo48.1% ±1.2$0.008356 s

90Claude Haiku 4.5 Thinking46.1% ±1.2$0.02740 s

91Llama 3.2 11B Vision Instruct Turbo38.8% ±1.2$0.0001738 s

92Qwen3.5 Plus Thinking22.8% ±1.0$0.0152 min