Top AI Models Flunk MazeBench, Best Score Just 8%

Top frontier models nearly all fail the MazeBench maze-solving benchmark, with GPT-6.1 Terra scoring best at just 8%, Claude Sonnet 5.5 at 4%, and Gemini 4 Hydrogen at only 1%.

2026-11-13 ~ 2026-11-13 · 2 related posts

• MazeBench: Top Models Score as Low as 1-8% on New Benchmark — patience_cave · 2026-11-13
• MazeBench shows GPT-6.1 Terra leading at 8% vs Gemini 4 Hydrogen's 1% — patience_cave · 2026-11-13