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.
Rebuild a mobile health dashboard from a design image as a real web page.
Comparison
Model’s work
Agent session
Fable 5.1Claude Code 2.1.28332 minhigheffort6.2Min130Kout$11.25API cost
GPT-6 AstraCodex CLI 0.157.010 minhigheffort480Kin17Kout$1.72API cost
Opus 5.5Claude Code 2.1.28321 minhigheffort14Min110Kout$6.09API cost
Sonnet 5.5Claude Code 2.1.28412 minhigheffort5.2Min79Kout$2.17API cost
GPT-6.1 SolCodex CLI 0.159.012 minhigheffort850Kin20Kout$0.39API cost
Grok 4.7Grok Build 1.0.4143 minhigheffort13Min160Kout$7.96API cost
Sonnet 5Claude Code 2.1.2832 minhigheffort110Kin12Kout$0.19API cost
GPT-6 SolCodex CLI 0.157.09 minhigheffort940Kin25Kout$0.54API cost
Grok 4.6Grok Build 1.0.4128 minhigheffort7Min82Kout$4.54API cost
Haiku 4.5Claude Code 2.1.2837 minhigheffort2.5Min42Kout$0.53API cost
GPT-6 LunaCodex CLI 0.157.03 minhigheffort210Kin10Kout<$0.01API cost
haiku-5.5Claude Code 2.1.2939 minhigheffort4.3Min110Kout
gemini-3.1-proAntigravity CLI 1.2.146 minhigheffort95Kin39Kout$0.58API cost
gemini-3.8-flashAntigravity CLI 1.2.141 h 0 minhigheffort71Min410Kout$8.79API cost
gemini-3.6-flashAntigravity CLI 1.2.1421 minhigheffort7.8Min110Kout$1.29API cost