benchgap
Vision & documents

VideoMMMU leaderboard

As of 2026-11-22, the highest measured score on VideoMMMU is 88.7% by Qwen3.8 Max. 78 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.1 Terra89.2%estimated ± 2.4 pp, medium confidence
2Qwen3.8 Max88.7%measured
3Claude Mythos 588.2%estimated ± 1.5 pp, medium confidence
4Kimi K3.1 Pro88.1%estimated ± 1.5 pp, medium confidence
5GPT-5.6 Sol88.1%estimated ± 2.4 pp, medium confidence
6Glimpse88.0%estimated ± 1.5 pp, medium confidence
7Gemini 3 Pro87.6%measured
8Qwen3.8-Omni-Flash87.6%estimated ± 1.5 pp, medium confidence
9Kimi K387.6%estimated ± 1.5 pp, medium confidence
10Claude Opus 4.787.5%estimated ± 1.5 pp, medium confidence
11Qwen3.8-Flash-Next87.4%estimated ± 1.5 pp, medium confidence
12LongCat-2.587.3%estimated ± 1.5 pp, medium confidence
13Qwen3.8-27B87.3%estimated ± 1.5 pp, medium confidence
14Step 3.7 Flash87.2%estimated ± 10.3 pp, low confidence
15Claude Opus 4.887.2%estimated ± 1.5 pp, medium confidence
16GPT-5.587.2%estimated ± 2.4 pp, high confidence
17GLM-5.3-Flash87.1%estimated ± 1.5 pp, medium confidence
18Gemini 3.7 Flash86.9%estimated ± 1.5 pp, medium confidence
19Muse Spark Open86.9%estimated ± 10.3 pp, low confidence
20GPT-5.6 Terra86.9%estimated ± 2.4 pp, high confidence
21Muse Spark 1.186.9%estimated ± 1.5 pp, medium confidence
22Claude Sonnet 586.8%estimated ± 1.5 pp, medium confidence
23Inkling 2 Mini86.8%estimated ± 1.5 pp, medium confidence
24dots3-note Preview86.8%measured
25Inkling 2 Small86.7%estimated ± 1.5 pp, medium confidence
26Kimi K2.586.6%measured
27Sakana Fugu-Ultra86.5%estimated ± 1.5 pp, medium confidence
28Muse Spark86.4%estimated ± 1.5 pp, medium confidence
29Seed 2.1 Pro86.2%estimated ± 1.5 pp, medium confidence
30Sakana Fugu86.2%estimated ± 1.5 pp, medium confidence
31Gemini 3.5 Flash86.0%estimated ± 1.5 pp, medium confidence
32GPT-5.285.9%measured
33GPT-5.485.7%estimated ± 1.5 pp, medium confidence
34Hy485.7%estimated ± 2.4 pp, high confidence
35MiniMax M3.1 Flash85.7%estimated ± 1.5 pp, medium confidence
36Seed 2.1 Turbo85.7%estimated ± 1.5 pp, medium confidence
37GPT-5.6 Luna85.6%estimated ± 2.4 pp, high confidence
38Inkling85.6%estimated ± 1.5 pp, medium confidence
39Inkling-Small85.5%estimated ± 1.5 pp, medium confidence
40Grok 4.385.4%estimated ± 2.4 pp, high confidence
41MiMo-V2.585.4%estimated ± 1.5 pp, medium confidence
42Qwen3.7 Plus85.4%measured
43Nemotron 3.5 Super VL 120B A12B85.4%estimated ± 1.5 pp, medium confidence
44Pareto 26.985.4%estimated ± 2.4 pp, high confidence
45Kimi K2.685.3%estimated ± 1.5 pp, medium confidence
46Gemini 3.1 Pro85.3%estimated ± 1.5 pp, medium confidence
47Nemotron 3.5 Super VL 120B A12B NVFP485.3%estimated ± 1.5 pp, medium confidence
48Qwen3.5-27B85.2%estimated ± 1.5 pp, medium confidence
49Muse Glimmer 30B85.1%estimated ± 1.5 pp, medium confidence
50Claude Opus 4.684.9%estimated ± 2.4 pp, high confidence
51Qwen3.5-35B-A3B84.9%estimated ± 1.5 pp, medium confidence
52Claude Sonnet 4.684.9%estimated ± 1.5 pp, medium confidence
53Qwen3.5-122B-A10B84.9%estimated ± 1.5 pp, medium confidence
54Nemotron 3 Nano Omni 30B A3B84.8%estimated ± 1.5 pp, medium confidence
55Qwen3.5 397B84.7%measured
56Gemma 4 31B84.7%estimated ± 2.4 pp, high confidence
57MiniMax M384.6%measured
58GPT-5.4 mini84.5%estimated ± 2.4 pp, high confidence
59Gemini 3.1 Flash-Lite84.4%estimated ± 1.5 pp, medium confidence
60Claude Opus 4.584.4%measured
61Qwen3.6-27B84.4%measured
62Step 5 Preview84.1%estimated ± 2.4 pp, high confidence
63Qwen3.6 Plus84.0%measured
64Hy4 flash84.0%estimated ± 2.4 pp, high confidence
65Qwen3.6-35B-A3B83.7%measured
66Grok 4.2083.6%estimated ± 1.5 pp, low confidence
67Command A+83.3%estimated ± 1.5 pp, low confidence
68Gemma 4 26B A4B82.5%estimated ± 2.4 pp, high confidence
69Interfaze Beta80.4%estimated ± 2.4 pp, high confidence
70Gemma 4 12B78.6%estimated ± 2.4 pp, high confidence
71Claude 4 Opus Thinking76.3%estimated ± 2.3 pp, high confidence
72GPT-5.4 nano75.7%estimated ± 2.4 pp, high confidence
73GLM-4.6V74.7%measured
74Qwen3-VL-235B-A22B Instruct74.7%measured
75Claude 3.7 Sonnet Thinking74.4%estimated ± 2.3 pp, high confidence
76Claude 4 Opus72.8%estimated ± 2.3 pp, high confidence
77Gemini 2.5 Flash-Lite71.7%estimated ± 2.3 pp, high confidence
78Claude 3.7 Sonnet70.3%estimated ± 2.3 pp, high confidence
79GLM-4.6V-Flash70.1%measured
80Gemini 2.0 Flash66.8%estimated ± 2.3 pp, high confidence
81MiniMax-VL-0165.7%estimated ± 2.3 pp, high confidence
82Grok 262.4%estimated ± 2.3 pp, high confidence
83Gemini 2.0 Flash-Lite61.1%estimated ± 2.3 pp, high confidence
84Mistral Small 3.159.6%estimated ± 2.3 pp, high confidence
85Gemma 4 E4B58.9%estimated ± 2.4 pp, high confidence
86Grok 2 mini58.5%estimated ± 2.3 pp, high confidence
87Qwen2.5-VL-32B54.6%estimated ± 2.4 pp, high confidence
88Gemini 1.5 Flash52.4%estimated ± 2.3 pp, high confidence
89LFM3-VL-2B49.6%measured
90Gemma 4 E2B47.8%estimated ± 2.4 pp, medium confidence
91Ternary Bonsai 2 27B44.7%estimated ± 9.6 pp, low confidence
92LFM2.5-VL-3B43.6%estimated ± 2.3 pp, medium confidence
93ZAYA1-VL-8B42.3%estimated ± 2.3 pp, medium confidence
94LFM2.5-VL-450M38.7%estimated ± 2.3 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General