MathVision (MATH-Vision) — Explicit full subset (Kimi-VL A3B); not merged w/ plain mathvision given diverging medians.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Qwen2.5 VL 72B | Alibaba | 38.1 | 1 | 2026-06-05 |
| 2 | Kimi VL A3B Thinking | Moonshot | 36.8 | 1 | 2026-06-05 |
| 3 | Gemma 3 27B | 35.5 | 1 | 2026-06-05 | |
| 4 | Gemma 3 12B | 32.1 | 1 | 2026-06-05 | |
| 5 | GPT-4o | OpenAI | 30.4 | 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.