AdaptMMBench - Math: leaderboard
Metric: Accuracy (%) on the Math (without auxiliary-line problems) domain of AdaptMMBench, enhancement-transformed samples included, in adaptive mode, where the model decides whether to call visual tools (zoom-in, rotation, contrast) through code or function calls; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 87 | #77 |
| 2 | Qwen 3 VL 235B A22B Instruct | 76.5 | #264 |
| 3 | Qwen 3 VL 32B Instruct | 73.5 | #276 |
| 4 | GPT-5 | 71 | #91 |
| 5 | Qwen 3 VL 8B Instruct | 62 | #401 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=adaptmmbench-math · How It Works · Data refreshed daily, snapshot 2026-10-11.