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
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 system | Score |
|---|---|---|
| 1 | MiniMax M3· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 85.4% |
| 2 | Gemini 3.8 FlashAgent mode · Qwen Code· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.7% |
| 3 | Qwen3.8 Omni FlashAgent mode · Qwen Code· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 71.3% |
| 4 | Gemini 3.8 FlashStatic· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 71% |
| 5 | Qwen3.8 MaxQwen3.8-Max release vision/video table · subtitles enabled· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 68.3% |
| 6 | Qwen3.8 Omni FlashStatic· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 65% |
| 7 | Muse Spark Open 816B A42B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.5% |
| 8 | Kimi K2.5· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.1% |
| 9 | Qwen3.5 397B A17B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 55.9% |
| 10 | Gemini 3 Pro Previeww. sub @ 1fps· Video-MME-v2 | 49.4% |
| 11 | Qwen3.5 Omni PlusStatic· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 47.9% |
| 12 | Doubao Seed 2.0 Prow. sub @ 1fps· Video-MME-v2 | 43.3% |
| 13 | Gemini 3 Flashw. sub @ 1fps· Video-MME-v2 | 42.5% |
| 14 | Kimi K2.5w. sub @ 512· Video-MME-v2 | 40.7% |
| 15 | Qwen3.5 397B A17Bw. sub @ 512, think· Video-MME-v2 | 39.1% |
| 16 | MiMo V2 Omniw. sub @ 1fps· Video-MME-v2 | 38.6% |
| 17 | Gemini 3 Pro Previewwo sub @ 1fps· Video-MME-v2 | 38.2% |
| 18 | GPT 5w. sub @ 50· Video-MME-v2 | 37% |
| 19 | Kimi K2.5w. sub @ 64· Video-MME-v2 | 36.1% |
| 20 | Doubao Seed 2.0 Prowo sub @ 1fps· Video-MME-v2 | 35.2% |
| 21 | Kimi K2.5wo sub @ 512· Video-MME-v2 | 33% |
| 22 | Gemini 3 Flashwo sub @ 1fps· Video-MME-v2 | 32.9% |
| 23 | Qwen3.5 27Bw. sub @ 512, think· Video-MME-v2 | 31.4% |
| 24 | Qwen3.5 397B A17Bw. sub @ 64, think· Video-MME-v2 | 30.6% |
| 25 | Qwen3.5 397B A17Bwo sub @ 512, think· Video-MME-v2 | 30.3% |
| 26 | MiMo V2 Omniwo sub @ 1fps· Video-MME-v2 | 29.9% |
| 27 | Qwen3.5 27Bw. sub @ 64, think· Video-MME-v2 | 29.6% |
| 28 | Gemma 4 31B ITw. sub @ 60· Video-MME-v2 | 28.2% |
| 29 | Qwen3 VL 235Bw. sub @ 512, think· Video-MME-v2 | 28.1% |
| 30 | Qwen3.5 Omni Plusw. sub @ 512· Video-MME-v2 | 27.4% |
| 31 | Kimi K2.5wo sub @ 64· Video-MME-v2 | 27.4% |
| 32 | Qwen3.5 122B-A10Bw. sub @ 64, think· Video-MME-v2 | 26.7% |
| 33 | GPT 5wo sub @ 50· Video-MME-v2 | 26.4% |
| 34 | Qwen3 VL 235Bw. sub @ 64, think· Video-MME-v2 | 26.3% |
| 35 | Qwen3.5 9Bw. sub @ 512, think· Video-MME-v2 | 26.2% |
| 36 | Qwen3.5 35B A3Bw. sub @ 64, think· Video-MME-v2 | 25.8% |
| 37 | Qwen3.5 Omni Plusw. sub @ 64· Video-MME-v2 | 25.6% |
| 38 | Qwen3.5 122B-A10Bw. sub @ 512, think· Video-MME-v2 | 25.1% |
| 39 | Qwen3 VL 235Bw. sub @ 64, instruct· Video-MME-v2 | 25% |
| 40 | Qwen3.5 397B A17Bw. sub @ 64, instruct· Video-MME-v2 | 24.5% |
| 41 | Qwen3.5 27Bw. sub @ 64, instruct· Video-MME-v2 | 23.9% |
| 42 | Qwen3.5 35B A3Bw. sub @ 512, think· Video-MME-v2 | 23.9% |
| 43 | Qwen3.5 9Bw. sub @ 64, think· Video-MME-v2 | 23.2% |
| 44 | internvl3-5-241b-a28bw. sub @ 64, instruct· Video-MME-v2 | 23.1% |
| 45 | Qwen2.5 VL 72B Instructw. sub @ 64, instruct· Video-MME-v2 | 22.1% |
| 46 | qwen3-vl-30bw. sub @ 512, think· Video-MME-v2 | 22.1% |
| 47 | GLM 4.5Vw. sub @ 64, think· Video-MME-v2 | 21.8% |
| 48 | qwen3-vl-30bw. sub @ 64, think· Video-MME-v2 | 21.1% |
| 49 | Qwen3.5 122B-A10Bw. sub @ 64, instruct· Video-MME-v2 | 20.9% |
| 50 | Qwen3.5 397B A17Bwo sub @ 64, think· Video-MME-v2 | 20.7% |
Showing 1–50 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 system | Score |
|---|---|
| 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