OpenVLM MMStar - Instance Reasoning: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 82.8 |
| 2 | GPT-5 | 79.6 |
| 3 | GPT-5 Nano | 78.8 |
| 4 | GPT-5 Mini (2025-08-07) | 78.4 |
| 5 | InternVL3-78B | 78 |
| 6 | InternVL3-38B | 75.2 |
| 7 | InternVL2.5-78B | 75.2 |
| 8 | InternVL3-8B | 74.8 |
| 9 | GPT-4.1 (2025-04-14) | 74.4 |
| 10 | Gemini 2.0 Flash | 74 |
| 11 | Qwen 2 VL 72B | 72.4 |
| 12 | Gemini 1.5 Pro (002) | 72.4 |
| 13 | Claude 3.7 Sonnet | 72 |
| 14 | GPT-4o ChatGPT | 71.6 |
| 15 | Gemini 1.5 Flash (002) | 70.8 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmstar-instance-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-05.