PAGE Bench: leaderboard
Precise geometric construction in the GeoGebra drawing interface (PAGE Bench, from the PAGER paper by Wei et al., 2026): a vision-language agent reads a school geometry problem and draws the construction with click, paint and type actions, where a painted point counts only within 5 pixels of the reference, on 463 test problems. The Overall Score is the mean of the Middle Process Score and the Final Result Score, on a 0-100 scale; higher is better.
Metric: Overall Score (0-100; mean of the middle-process and final-result scores). Source: arxiv.org. Saturation forecast: Around March 2028. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 24.36 | #91 |
| 2 | Qwen 3.6 Plus | 19.56 | #217 |
| 3 | Claude Sonnet 4.6 | 14.91 | #158 |
| 4 | GPT-5.4 | 14.28 | #121 |
| 5 | GLM-4.5V | 9.37 | #739 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=page-bench · How It Works · Data refreshed daily, snapshot 2026-11-12.