Pause-and-Think-B - Goal Planning: leaderboard
Metric: Binary validity accuracy (%) on the 200 goal-planning items (next-step action suggestion), GPT-5.1 judge, three runs; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 86.52 |
| 2 | Qwen 3 VL 235B A22B Instruct | 72.26 |
| 3 | Gemini 2.5 Pro | 69.85 |
| 4 | GPT-4o | 68.13 |
| 5 | Qwen 3 VL 4B Instruct | 53.76 |
Interactive version: theaggregate.ai/benchmark?slug=pause-and-think-b-goal-planning · How It Works · Data refreshed daily, snapshot 2026-09-29.