Pencil Puzzle Bench - Yajilin: leaderboard
Metric: Direct-ask Success Rate (%). Source: ppbench.com. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (High) | 20 |
| 2 | Gemini 3.1 Pro (Preview) | 13.3 |
| 3 | GPT-5.2 (xHigh) | 13.3 |
| 4 | Gemini 3 Pro | 6.7 |
| 5 | Claude Opus 4.6 (Thinking) | 6.7 |
| 6 | GPT-4o | 0 |
| 7 | Gemini 2.5 Pro | 0 |
| 8 | Claude Sonnet 4.6 | 0 |
| 9 | GPT-4.1 | 0 |
| 10 | GPT-5.2 | 0 |
| 11 | Claude Sonnet 4.5 | 0 |
| 12 | GPT-OSS-120B | 0 |
| 13 | Claude Opus 4.6 | 0 |
| 14 | Kimi K2.5 | 0 |
| 15 | DeepSeek V3.2 | 0 |
Interactive version: theaggregate.ai/benchmark?slug=pencil-puzzle-bench-yajilin · How It Works · Data refreshed daily, snapshot 2026-09-05.