MMStar: leaderboard
MMStar evaluates vision-indispensable multimodal capabilities across perception, reasoning, STEM, and math axes.
Metric: Accuracy (%). Source: huggingface.co. Status: saturation imminent. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 75.7 |
| 2 | Gemini 2.5 Pro | 73.6 |
| 3 | InternVL3-78B | 73.4 |
| 4 | GPT-5 Mini (2025-08-07) | 73.2 |
| 5 | InternVL3-38B | 72.6 |
| 6 | Doubao-1.5-Pro | 71.9 |
| 7 | GPT-4o ChatGPT | 70.2 |
| 8 | GPT-4.1 (2025-04-14) | 69.8 |
| 9 | InternVL2.5-78B | 69.5 |
| 10 | Gemini 2.0 Flash | 69.4 |
| 11 | GPT-4.5 | 69.3 |
| 12 | InternVL3-8B | 68.7 |
| 13 | Qwen 2 VL 72B | 68.6 |
| 14 | GPT-5 Nano | 68.3 |
| 15 | Gemini 1.5 Pro (002) | 67.1 |
Interactive version: theaggregate.ai/benchmark?slug=mmstar · How It Works · Data refreshed daily, snapshot 2026-09-05.