OpenVLM Video: leaderboard
OpenCompass OpenVLM video leaderboard aggregating video-language model results on Video-MME, MVBench, MMBench-Video, TempCompass, and MLVU.
Metric: Average Normalized Score (%). Source: huggingface.co. Status: saturated. 49 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-78B | 61.5 |
| 2 | InternVL3-38B | 60.74 |
| 3 | Qwen 2 VL 72B | 57.55 |
| 4 | InternVL2.5-78B | 56.68 |
| 5 | InternVL3-8B | 56.4 |
| 6 | Gemini 2.0 Flash | 53.34 |
| 7 | GPT-4o (2024-08-06) | 53.23 |
| 8 | Qwen 2 VL 7B | 51.9 |
| 9 | InternVL2-8B | 48.14 |
| 10 | Gemini 1.5 Flash | 48.1 |
| 11 | Gemini 1.5 Pro | 38.8 |
| 12 | Claude 3.5 Sonnet | 38.76 |
| 13 | Phi-4 Multimodal Instruct | 35.99 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-video · How It Works · Data refreshed daily, snapshot 2026-09-05.