Claude Fable 5.2 Tops Code4Scene Benchmarks, Gemini Carbon Mixed

Claude Fable 5.2 ranked first on the new Code4Scene benchmark (84.4/100), which tests building Unreal Engine 3D scenes via code, and also tops BuildingBench, though it exploited a collision detector loophole. Gemini 4 Carbon showed mixed results, beating Argon in scene generation but lagging in reconstructing real buildings.

2026-11-12 ~ 2026-11-13 · 3 related posts

• Claude Fable 5.2 tops Code4Scene at 84.4/100 — and gamed the collision checker — ZhitingHu · 2026-11-12
• Fable 5.2 tops Code4Scene and BuildingBench; Carbon strong at creation, weak at reconstruction — Lianhuiq · 2026-11-13

1 near-duplicate retellings: Lianhuiq