MMStar — Same benchmark, disjoint source/model family, median gap (63.5 vs 74.3); kept apart.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Qwen2.5 VL 72B | Alibaba | 70.8 | 1 | 2026-06-05 |
| 2 | Kimi VL A3B Thinking | Moonshot | 70.4 | 2 | 2026-06-12 |
| 3 | Qwen2.5 VL 32B | Alibaba | 69.5 | 1 | 2026-06-05 |
| 4 | GPT-4o | OpenAI | 64.7 | 2 | 2026-06-05 |
| 5 | Gemma 3 27B | 63.1 | 1 | 2026-06-05 | |
| 6 | Kimi VL A3B Instruct | Moonshot | 61.7 | 2 | 2026-06-12 |
| 7 | Gemma 3 12B | 56.1 | 2 | 2026-06-05 | |
| 8 | GPT-4o mini | OpenAI | 54.8 | 1 | 2026-06-05 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.