OpenVLM MMT-Bench - Behavior Anomaly Detection: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.7 Sonnet | 80 |
| 2 | Claude 3 Haiku | 75 |
| 3 | Pixtral-12B | 70 |
| 4 | Claude 3.5 Sonnet | 70 |
| 5 | Aquila-VL-2B | 65 |
| 6 | Grok 2 (1212) | 65 |
| 7 | GPT-4.1 Mini | 60 |
| 8 | InternVL3-78B | 60 |
| 9 | Qwen 2 VL 72B | 60 |
| 10 | Qwen 2 VL 7B | 55 |
| 11 | InternVL3-38B | 55 |
| 12 | Gemini 1.5 Pro | 55 |
| 13 | GPT-4.1 Nano | 45 |
| 14 | InternVL3-8B | 40 |
| 15 | InternVL2-8B | 40 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-behavior-anomaly-detection · How It Works · Data refreshed daily, snapshot 2026-09-05.