ZebraLogic: leaderboard
Logic grid puzzle benchmark from Allen AI testing systematic constraint satisfaction and deductive reasoning across varying difficulty levels.
Metric: Puzzle Accuracy (%). Source: huggingface.co. Status: saturated. 62 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 Mini (2025-01-31) (High) | 91.7 |
| 2 | O3 Mini (Medium) | 88.9 |
| 3 | O1 (2024-12-17) | 81 |
| 4 | DeepSeek R1 | 78.7 |
| 5 | O3 Mini (Low) | 74.8 |
| 6 | O1 Preview (2024-09-12) | 71.4 |
| 7 | O1 Mini (2024-09-12) | 59.7 |
| 8 | DeepSeek V3 | 42.1 |
| 9 | Claude 3.5 Sonnet (20241022) | 36.2 |
| 10 | Claude 3.5 Sonnet (20240620) | 33.4 |
| 11 | GPT-4o (2024-08-06) | 31.7 |
| 12 | Mistral Large 2 (Jul) | 29 |
| 13 | GPT-4 Turbo | 28.4 |
| 14 | GPT-4o (2024-05-13) | 28.2 |
| 15 | Grok 2 (1212) | 27.7 |
Interactive version: theaggregate.ai/benchmark?slug=zebralogic · How It Works · Data refreshed daily, snapshot 2026-09-05.