OpenVLM MMStar - Instance Reasoning: leaderboard
OpenCompass OpenVLM MMStar subcategory score for visual instance reasoning questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 82.8 | #152 |
| 2 | GPT-5 | 79.6 | #98 |
| 3 | GPT-5 Nano | 78.8 | #425 |
| 4 | GPT-5 Mini (2025-08-07) | 78.4 | #172 |
| 5 | InternVL3-78B | 78 | #352 |
| 6 | InternVL3-38B | 75.2 | #405 |
| 7 | InternVL2.5-78B | 75.2 | #431 |
| 8 | InternVL3-8B | 74.8 | #620 |
| 9 | GPT-4.1 (2025-04-14) | 74.4 | #250 |
| 10 | Gemini 2.0 Flash | 74 | #342 |
| 11 | Gemini 1.5 Pro (002) | 72.4 | #349 |
| 12 | Qwen 2 VL 72B | 72.4 | #460 |
| 13 | InternVL3-14B | 72 | #505 |
| 14 | Claude 3.7 Sonnet | 72 | #249 |
| 15 | GPT-4o ChatGPT | 71.6 |
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-instance-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.