OpenVLM Video - MVBench - Unexpected Action: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 37 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2 VL 72B | 87.5 |
| 2 | InternVL3-78B | 87 |
| 3 | GPT-4o (2024-08-06) | 82 |
| 4 | Gemini 2.0 Flash | 81.5 |
| 5 | InternVL2.5-78B | 81.5 |
| 6 | Qwen 2 VL 7B | 77 |
| 7 | InternVL3-8B | 76.5 |
| 8 | Gemini 1.5 Flash | 75 |
| 9 | Phi-4 Multimodal Instruct | 73 |
| 10 | InternVL2-8B | 69.5 |
| 11 | Claude 3.5 Sonnet | 40.5 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-video-mvbench-unexpected-action · How It Works · Data refreshed daily, snapshot 2026-09-19.