OpenVLM MMT-Bench - MeViS: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemma 3 12B | 50 |
| 2 | GPT-4.1 (2025-04-14) | 50 |
| 3 | InternVL3-78B | 40 |
| 4 | InternVL3-38B | 40 |
| 5 | Aquila-VL-2B | 40 |
| 6 | Claude 3.7 Sonnet | 40 |
| 7 | Grok 2 (1212) | 40 |
| 8 | Claude 3 Haiku | 40 |
| 9 | Llama 3.2 90B Vision Instruct | 40 |
| 10 | Qwen 2 VL 72B | 35 |
| 11 | Qwen 2 VL 7B | 35 |
| 12 | Qwen 2 VL 2B | 35 |
| 13 | GPT-4.1 Nano | 35 |
| 14 | InternVL2-8B | 30 |
| 15 | Pixtral-12B | 30 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-mevis · How It Works · Data refreshed daily, snapshot 2026-09-05.