OpenVLM MMBench V1.1 EN - Physical Relation: leaderboard
OpenCompass OpenVLM MMBench V1.1 English split accuracy for physical relation reasoning.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3-78B | 77.3 | #352 |
| 2 | InternVL2.5-78B | 76 | #431 |
| 3 | InternVL3-38B | 74.7 | #405 |
| 4 | GPT-5 | 73.3 | #98 |
| 5 | InternVL3-8B | 72 | #620 |
| 6 | GPT-5 Mini (2025-08-07) | 72 | #172 |
| 7 | Gemini 2.5 Pro | 70.7 | #152 |
| 8 | InternVL3-14B | 70.7 | #505 |
| 9 | InternVL2-8B | 65.3 | #817 |
| 10 | GPT-4.1 Mini | 64 | #355 |
| 11 | MiniCPM-V-2.6 | 64 | #836 |
| 12 | GPT-4.1 (2025-04-14) | 62.7 | #250 |
| 13 | GPT-4.5 | 62.7 | #274 |
| 14 | Llama 3.2 90B Vision Instruct | 62.7 | #622 |
| 15 | GPT-4o ChatGPT | 62.7 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmbench-v1-1-en-physical-relation · How It Works · Data refreshed daily, snapshot 2026-10-20.