OpenVLM MMStar - Fine-Grained Perception: leaderboard
OpenCompass OpenVLM MMStar subcategory score for fine-grained visual perception questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2028. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 66.8 | #152 |
| 2 | InternVL3-38B | 64.4 | #405 |
| 3 | InternVL3-78B | 64 | #352 |
| 4 | Qwen 2 VL 72B | 61.2 | #460 |
| 5 | InternVL3-8B | 60 | #620 |
| 6 | InternVL3-14B | 60 | #505 |
| 7 | GPT-5 | 58.8 | #98 |
| 8 | GPT-5 Mini (2025-08-07) | 58.8 | #172 |
| 9 | GPT-4.1 Mini | 58.4 | #355 |
| 10 | GPT-4.5 | 58.4 | #274 |
| 11 | Gemini 2.0 Flash | 57.6 | #342 |
| 12 | Qwen 2.5 VL 7B | 57.6 | #643 |
| 13 | Gemini 1.5 Pro (002) | 57.2 | #349 |
| 14 | GPT-4o ChatGPT | 56.8 | |
| 15 | Qwen 2 VL 7B | 56.4 | #750 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmstar-fine-grained-perception · How It Works · Data refreshed daily, snapshot 2026-10-20.