SudokuBench (Multi-step, 4x4) — leaderboard
SudokuBench multi-step mode on 4x4 grids. Smaller grids test basic constraint propagation with iterative solving.
Metric: 4x4 Solve Rate (%). Source: pub.sakana.ai. Status: saturation imminent. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 (High) | 80 |
| 2 | Gemini 2.5 Pro | 73.33 |
| 3 | GPT-5 (Medium) | 66.7 |
| 4 | DeepSeek R1 | 60 |
| 5 | O3 Mini (High) | 60 |
| 6 | Grok 3 Mini | 53.33 |
| 7 | GPT-5 (Low) | 53.3 |
| 8 | Qwen 3 235B A22B | 40 |
| 9 | Claude 3.7 Sonnet (Thinking) | 40 |
| 10 | QwQ-32B | 26.67 |
| 11 | GPT-4.1 | 13.33 |
| 12 | Qwen 3 30B A3B | 6.67 |
| 13 | Llama 4 Maverick | 0 |
| 14 | Llama 4 Scout | 0 |
Interactive version: theaggregate.ai/benchmark?slug=sudokubench-multi-step-4x4 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.