PAGE Bench - Middle Process Score: leaderboard
Process score of PAGE Bench (Wei et al., 2026): the agent's GeoGebra action sequence for each of 463 geometry construction problems is compared step by step with a verified reference trajectory. The score weights full task success 0.6, step success 0.2, parameter accuracy 0.1 and action-type accuracy 0.1, on a 0-100 scale; higher is better.
Metric: Middle Process Score (0-100; 0.6 task success + 0.2 step success + 0.1 parameter accuracy + 0.1 action-type accuracy). Source: arxiv.org. Saturation forecast: Around August 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 32.41 | #91 |
| 2 | Qwen 3.6 Plus | 27.41 | #217 |
| 3 | Claude Sonnet 4.6 | 21.17 | #158 |
| 4 | GPT-5.4 | 18.59 | #121 |
| 5 | GLM-4.5V | 11.46 | #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-middle-process-score · How It Works · Data refreshed daily, snapshot 2026-11-12.