NonoBench — leaderboard
Nonogram puzzle solving benchmark testing LLM spatial reasoning and constraint satisfaction through Japanese logic puzzles.
Metric: Overall Accuracy (%). Source: www.nonobench.com. Status: saturation imminent. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (xHigh) | 63.3 |
| 2 | Gemini 3.1 Pro (Preview) (High) | 63.3 |
| 3 | GPT-5.4 (High) | 60 |
| 4 | Claude Opus 4.5 (High) | 56.7 |
| 5 | GPT-5.2 (High) | 53.3 |
| 6 | Gemini 3 Pro (Preview) (High) | 53.3 |
| 7 | GPT-5.2 (xHigh) | 50 |
| 8 | GPT-5.4 (Low) | 46.7 |
| 9 | GPT-5.2 (Low) | 46.7 |
| 10 | Gemini 3 Flash (Preview) (High) | 46.7 |
| 11 | Gemini 3.1 Pro (Preview) (Low) | 40 |
| 12 | GPT-OSS-120B (High) | 36.7 |
| 13 | DeepSeek V3.2 Speciale | 36.7 |
| 14 | DeepSeek V3.2 (High) | 36.7 |
| 15 | Grok 4 | 33.3 |
Interactive version: theaggregate.ai/benchmark?slug=nonobench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.