Fable 5.2 Tops MazeBench With 21% After 90 Hours and $6,000 in Credits

patience_cave · x · 2026-11-16

A new high score on the third-party MazeBench benchmark: after 90 hours and $6,000 in credits spent navigating its monster 3D environment, Fable 5.2 (Claude Code max) scored 21%, vaulting over GPT-6 Astra's 16%.

Full leaderboard: Fable 5.2 at 21%, GPT-6 Astra at 16%, GPT-6.1 Sol / Claude Opus 5.5 / Gemini 4 Argon at 9% each, GPT-6.1 Terra at 7%, Gemini 4 Carbon at 6%, Claude Sonnet 5.5 at 4%.

Original post →

More from Models

Models channel →