OpenVLM MMT-Bench - Human Keypoint Detection: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2 VL 72B | 100 |
| 2 | InternVL3-38B | 100 |
| 3 | Grok 2 (1212) | 95 |
| 4 | InternVL3-78B | 90 |
| 5 | InternVL2-8B | 80 |
| 6 | GPT-4.1 (2025-04-14) | 80 |
| 7 | Claude 3 Haiku | 80 |
| 8 | InternVL3-8B | 70 |
| 9 | GPT-4.1 Mini | 65 |
| 10 | Gemma 3 27B | 65 |
| 11 | Gemini 1.5 Pro | 60 |
| 12 | Gemma 3 12B | 60 |
| 13 | Aquila-VL-2B | 55 |
| 14 | GPT-4.1 Nano | 55 |
| 15 | Gemma 3 4B | 55 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-human-keypoint-detection · How It Works · Data refreshed daily, snapshot 2026-09-05.