Pencil Puzzle Bench - Mashu: leaderboard
Models draw a single Mashu loop through black and white pearls, applying each pearl's turn or straight-line rule correctly. Direct-ask exact success fraction for 15fixed golden-300 puzzles of this type; archived DirectAskStrategy_05aaaac9, latest record per source model and puzzle. No agentic or revised strategy cohort mixed.
Metric: Direct-ask Success Rate (%). Source: ppbench.com. Saturation forecast: Around January 2027. 51 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 60 | #63 (Claude Opus 4.6) |
| 2 | GPT-5.2 (xHigh) | 60 | #110 (GPT-5.2) |
| 3 | GPT-5.2 (High) | 33.33 | #110 (GPT-5.2) |
| 4 | Gemini 3.1 Pro (Preview) | 20 | #58 |
| 5 | GPT-5.2 (Medium) | 13.33 | #110 (GPT-5.2) |
| 6 | DeepSeek V3.2 | 6.67 | #205 |
| 7 | Gemini 3 Pro | 6.67 | #80 |
| 8 | Grok 4.1 Fast | 6.67 | #217 |
| 9 | Claude Opus 4.5 (Thinking) | 6.67 | #83 (Claude Opus 4.5) |
| 10 | Claude Sonnet 4.6 (Thinking) | 6.67 | #90 (Claude Sonnet 4.6) |
| 11 | GPT-5.1 (Medium) | 6.67 | #135 (GPT-5.1) |
| 12 | Gemini 3 Flash (High) | 6.67 | #99 (Gemini 3 Flash) |
| 13 | GPT-5.2 Pro | 6.67 | #61 |
| 14 | Gemini 3 Flash (Minimal) | 6.67 | #99 (Gemini 3 Flash) |
| 15 | GPT-4o | 0 | #343 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pencil-puzzle-bench-mashu · How It Works · Data refreshed daily, snapshot 2026-10-20.