OpenVLM Video - MLVU - Topic Reasoning: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 37 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-78B | 90.9 |
| 2 | Gemini 2.0 Flash | 89.4 |
| 3 | InternVL3-8B | 89.4 |
| 4 | InternVL2.5-78B | 89.4 |
| 5 | Qwen 2 VL 72B | 88.2 |
| 6 | GPT-4o (2024-08-06) | 85.9 |
| 7 | Qwen 2 VL 7B | 84.8 |
| 8 | Claude 3.5 Sonnet | 83.7 |
| 9 | InternVL2-8B | 82.9 |
| 10 | Gemini 1.5 Flash | 82.5 |
| 11 | Phi-4 Multimodal Instruct | 60.8 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-video-mlvu-topic-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.