OpenVLM MMT-Bench - Reasoning Segmentation: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-38B | 89.5 |
| 2 | InternVL3-78B | 78.9 |
| 3 | Qwen 2 VL 72B | 78.9 |
| 4 | InternVL3-8B | 73.7 |
| 5 | Qwen 2 VL 7B | 73.7 |
| 6 | Grok 2 (1212) | 68.4 |
| 7 | Gemini 1.5 Pro | 63.2 |
| 8 | InternVL2-8B | 57.9 |
| 9 | Aquila-VL-2B | 57.9 |
| 10 | Claude 3.7 Sonnet | 57.9 |
| 11 | GPT-4.1 Mini | 52.6 |
| 12 | Pixtral-12B | 52.6 |
| 13 | GPT-4.1 (2025-04-14) | 52.6 |
| 14 | GPT-4.1 Nano | 47.4 |
| 15 | Llama 3.2 11B Instruct | 47.4 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-reasoning-segmentation · How It Works · Data refreshed daily, snapshot 2026-09-05.