ReasoningEst. 2024

ZebraLogic

ZebraLogic tests logical deduction ability using Zebra puzzles (also known as Einstein's riddle). Models must use constraint satisfaction and logical elimination to solve grid-based logic puzzles of increasing complexity.

Metrics

Solve rate (%) on logic puzzles

Created By

Allen Institute for AI

Top Model Scores

RankModelScoreDate
1Claude Opus 4.674.8%2026-02
2GPT-5.272.3%2026-03
3Gemini 3 Ultra68.9%2026-01
4Grok 464.1%2026-02
5DeepSeek V359.7%2026-01