Benchmark
Video-MME
Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.
- Models measured
- 35
- Scores recorded
- 44
- Sources
- 3
- IQ weight
- 2.35
Score distribution
- 62%
- Mean
- 83.7%
- Median
- 36.1%
- Std dev
- 0.639%
- 10th percentile
- 88%
- 90th percentile
- 0.513%
- Min
This benchmark has a weight of 2.35 for chat models, measured over 33 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | 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. | 90.4% |
| 2 | Kimi K3with subtitles · three-run average · max effort · temp 1.0 · top_p .95· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90% |
| 3 | Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.2% |
| 4 | Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89% |
| 5 | Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88% |
| 6 | Qwen3.7 PlusQwen3.7-Plus release · subtitles enabled· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 88% |
| 7 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise · with subtitles· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.8% |
| 8 | MiMo V2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.7% |
| 9 | MiMo V2.5Xiaomi MiMo-V2.5 official release/model card · subtitle/audio setting unspecified· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.7% |
| 10 | Qwen3.6 27BQwen official release table · thinking mode unless benchmark requires otherwise · with subtitles· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.7% |
| 11 | Qwen3.5 397B A17BQwen3.5 release/model card · with subtitles· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.5% |
| 12 | Kimi K2.5setting not named by the source· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.4% |
| 13 | Kimi K2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.4% |
| 14 | Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash · with subtitles· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 86.6% |
| 15 | MiniMax M3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 85.4% |
| 16 | MiniMax M3with subtitles · 1 FPS · ≤1024 frames· Reported by MiniMax Research· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.4% |
| 17 | MiMo V2 OmniXiaomi MiMo-V2-Omni release table · native audio-video input · exact benchmark version unspecified· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.3% |
| 18 | LongCat 2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 85.2% |
| 19 | Gemini 2.5 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.8% |
| 20 | Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.2% |
| 21 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise · without subtitles· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.2% |
| 22 | Qwen3.5 397B A17BQwen3.5 release/model card · without subtitles· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.7% |
| 23 | Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash · without subtitles· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.5% |
| 24 | Qwen3.5 Omni PlusQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 81.9% |
| 25 | Gemini 1.5 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.6% |
| 26 | Qwen3.5 Omni FlashQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 77% |
| 27 | Gemini 1.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.1% |
| 28 | gemini-1.5-pro-001· Epoch AI Benchmarking Hub data | 75% |
| 29 | qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 74.5% |
| 30 | Qwen2.5 VL 72B Instruct· Epoch AI Benchmarking Hub data | 73.5% |
| 31 | qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 73.3% |
| 32 | GPT 4o· Epoch AI Benchmarking Hub data | 71.9% |
| 33 | Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 71.8% |
| 34 | Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 71.4% |
| 35 | Qwen2-VL (72B) Instruct· Epoch AI Benchmarking Hub data | 71.2% |
| 36 | gemini-1.5-flash-001· Epoch AI Benchmarking Hub data | 70.3% |
| 37 | Gemini 1.5 Flash-8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 66.2% |
| 38 | GPT-4o mini· Epoch AI Benchmarking Hub data | 64.8% |
| 39 | Qwen2 VL 7b Instruct· Epoch AI Benchmarking Hub data | 63.9% |
| 40 | MiniCPM-V 2.6· Epoch AI Benchmarking Hub data | 60.9% |
| 41 | Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data | 60% |
| 42 | GPT-4-1106-vision-preview· Epoch AI Benchmarking Hub data | 59.9% |
| 43 | Phi 4 Multimodal Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 55% |
| 44 | Qwen-VL Max· Epoch AI Benchmarking Hub data | 51.3% |
Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats · 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.