Pencil Puzzle Bench - Mashu — leaderboard
Metric: Direct-ask Success Rate (%). Source: ppbench.com. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 60 |
| 2 | GPT-5.2 (xHigh) | 60 |
| 3 | GPT-5.2 (High) | 33.3 |
| 4 | Gemini 3.1 Pro (Preview) | 20 |
| 5 | GPT-5.2 (Medium) | 13.3 |
| 6 | DeepSeek V3.2 | 6.7 |
| 7 | Gemini 3 Pro | 6.7 |
| 8 | Grok 4.1 Fast | 6.7 |
| 9 | Claude Opus 4.5 (Thinking) | 6.7 |
| 10 | GPT-5.1 (Medium) | 6.7 |
| 11 | GPT-5.2 Pro | 6.7 |
| 12 | Gemini 3 Flash (High) | 6.7 |
| 13 | Claude Sonnet 4.6 (Thinking) | 6.7 |
| 14 | Gemini 2.5 Pro | 0 |
| 15 | GPT-4o | 0 |
Interactive version: theaggregate.ai/benchmark?slug=pencil-puzzle-bench-mashu · How the rankings work · Data refreshed daily, snapshot 2026-07-22.