Skip to content

Benchmark

Video-MME-v2

Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.

Models and systems measured
26
Scores recorded
89
Sources
3
IQ weight
1.91

Score distribution

28%
Mean
23.1%
Median
16.8%
Std dev
13.2%
10th percentile
55.9%
90th percentile
7.8%
Min
7.8%85.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.91 for chat models, measured over 25 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
51Qwen3.5 4Bw. sub @ 64, think· Video-MME-v220.7%
52Qwen: Qwen3 VL 8B Thinkingw. sub @ 512, think· Video-MME-v220.2%
53Qwen3.5 35B A3Bw. sub @ 64, instruct· Video-MME-v219.9%
54Qwen: Qwen3 VL 8B Thinkingw. sub @ 64, think· Video-MME-v219.7%
55Qwen3.5 27Bwo sub @ 512, think· Video-MME-v219.5%
56Qwen3.5 9Bwo sub @ 512, think· Video-MME-v219.5%
57Qwen3.5 Omni Pluswo sub @ 512· Video-MME-v219.3%
58Gemma 4 31B ITwo sub @ 60· Video-MME-v219%
59Qwen3 VL 235Bwo sub @ 512, think· Video-MME-v219%
60Qwen3.5 9Bw. sub @ 64, instruct· Video-MME-v218.3%
61Qwen: Qwen3 VL 8B Thinkingw. sub @ 64, instruct· Video-MME-v218.2%
62Qwen3.5 27Bwo sub @ 64, think· Video-MME-v217.6%
63Qwen3.5 Omni Pluswo sub @ 64· Video-MME-v217.3%
64Qwen3.5 397B A17Bwo sub @ 64, instruct· Video-MME-v216.9%
65qwen3-vl-30bw. sub @ 64, instruct· Video-MME-v216.8%
66Qwen3 VL 235Bwo sub @ 64, instruct· Video-MME-v216.5%
67Qwen3.5 122B-A10Bwo sub @ 64, think· Video-MME-v216.3%
68Qwen3 VL 235Bwo sub @ 64, think· Video-MME-v215.9%
69internvl3-5-241b-a28bwo sub @ 64, instruct· Video-MME-v215.8%
70Qwen3.5 35B A3Bwo sub @ 64, think· Video-MME-v215.6%
71Qwen3.5 4Bw. sub @ 64, instruct· Video-MME-v215.4%
72Qwen: Qwen3 VL 8B Thinkingwo sub @ 512, think· Video-MME-v214.4%
73Qwen3.5 122B-A10Bwo sub @ 512, think· Video-MME-v214%
74Qwen3.5 27Bwo sub @ 64, instruct· Video-MME-v214%
75Qwen2.5 VL 72B Instructwo sub @ 64, instruct· Video-MME-v213.7%
76Qwen3.5 9Bwo sub @ 64, think· Video-MME-v213.7%
77qwen3-vl-30bwo sub @ 64, think· Video-MME-v213.5%
78GLM 4.5Vwo sub @ 64, think· Video-MME-v213.3%
79qwen3-vl-30bwo sub @ 512, think· Video-MME-v213.2%
80Qwen3.5 122B-A10Bwo sub @ 64, instruct· Video-MME-v212.5%
81Qwen: Qwen3 VL 8B Thinkingwo sub @ 64, instruct· Video-MME-v212.4%
82Qwen3.5 35B A3Bwo sub @ 512, think· Video-MME-v212.2%
83Qwen: Qwen3 VL 8B Thinkingwo sub @ 64, think· Video-MME-v211.8%
84Qwen3.5 4Bwo sub @ 64, think· Video-MME-v211.6%
85Qwen3.5 35B A3Bwo sub @ 64, instruct· Video-MME-v211.3%
86Qwen3.5 9Bwo sub @ 64, instruct· Video-MME-v210.3%
87Qwen3.5 4Bwo sub @ 64, instruct· Video-MME-v29.5%
88qwen3-vl-30bwo sub @ 64, instruct· Video-MME-v27.8%

Showing 51–88 of 88 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Seed-2.0-LiteStatic· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.9%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Hugging Face official benchmark leaderboards; licence varies by benchmark