OpenVLM MMT-Bench - Action Quality Assessment: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-78B | 50 |
| 2 | InternVL3-8B | 45 |
| 3 | GPT-4.1 Nano | 30 |
| 4 | Gemini 1.5 Flash | 30 |
| 5 | Qwen 2 VL 72B | 25 |
| 6 | Qwen 2 VL 7B | 25 |
| 7 | Qwen 2 VL 2B | 25 |
| 8 | Gemini 1.5 Pro | 20 |
| 9 | Grok 2 (1212) | 20 |
| 10 | Llama 3.2 90B Vision Instruct | 20 |
| 11 | InternVL2-8B | 15 |
| 12 | InternVL3-38B | 15 |
| 13 | Pixtral-12B | 15 |
| 14 | Llama 3.2 11B Instruct | 15 |
| 15 | Gemma 3 27B | 15 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-action-quality-assessment · How It Works · Data refreshed daily, snapshot 2026-09-05.