User tests Fable 5, Opus 4.8, GPT-5.6 via playable 3D games

A Reddit user evaluated multiple frontier models including Fable 5, Opus 4.8, and GPT-5.6 using playable 3D games to assess spatial, temporal, and causal coherence. The benchmark aims to test world model capabilities.