Is it nerfed? We regularly give every model the same coding tasks, track whether its work gets worse, and let you know when it does.
Build an Angry Birds–style slingshot game with its own physics engine, playable on desktop and phone.
Desktop
Agent session
Fable 5.1Claude Code 2.1.28354 minhigheffort3.5Min160Kout$11.26API cost
GPT-6 AstraCodex CLI 0.157.024 minhigheffort1.4Min43Kout$4.08API cost
Opus 5.5Claude Code 2.1.28332 minhigheffort12Min170Kout$6.95API cost
Sonnet 5.5Claude Code 2.1.28456 minhigheffort31Min290Kout$9.97API cost
GPT-6.1 SolCodex CLI 0.159.028 minhigheffort1.7Min53Kout$0.85API cost
Grok 4.7Grok Build 1.0.411 h 2 minhigheffort8.9Min250Kout$6.40API cost
Sonnet 5Claude Code 2.1.28346 minhigheffort22Min220Kout$7.11API cost
GPT-6 SolCodex CLI 0.157.018 minhigheffort3Min48Kout$1.22API cost
Grok 4.6Grok Build 1.0.4135 minhigheffort3.6Min120Kout$2.93API cost
Haiku 4.5Claude Code 2.1.28310 minhigheffort9.3Min59Kout$1.33API cost
GPT-6 LunaCodex CLI 0.157.012 minhigheffort1Min38Kout$0.03API cost
haiku-5.5Claude Code 2.1.29338 minhigheffort10Min280Kout
gemini-3.6-flashAntigravity CLI 1.2.1412 minhigheffort2.7Min76Kout$1.12API cost
gemini-3.8-flashAntigravity CLI 1.2.1413 minhigheffort9.9Min140Kout$1.58API cost
gemini-3.1-proAntigravity CLI 1.2.145 minhigheffort540Kin36Kout$0.70API cost