NerfTrackerHow we test

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 a 3D kart racer in three.js that plays on desktop and phone.

Sort by
Desktop
Phone
Agent session
Fable 5.1Claude Code 2.1.28240 minhigheffort
GPT-6 AstraCodex CLI 0.156.123 minhigheffort
Opus 5.5Claude Code 2.1.28543 minhigheffort
Sonnet 5.5Claude Code 2.1.28427 minhigheffort
GPT-6.1 SolCodex CLI 0.159.033 minhigheffort
Grok 4.7Grok Build 1.0.4154 minhigheffort
Sonnet 5Claude Code 2.1.28233 minhigheffort
GPT-6 SolCodex CLI 0.156.114 minhigheffort
Grok 4.6Grok Build 1.0.4128 minhigheffort
Haiku 4.5Claude Code 2.1.2829 minhigheffort
GPT-6 LunaCodex CLI 0.157.010 minhigheffort
gemini-3.8-flashAntigravity CLI 1.2.1414 minhigheffort
haiku-5.5Claude Code 2.1.29324 minhigheffort
gemini-3.1-proAntigravity CLI 1.2.145 minhigheffort
gemini-3.6-flashAntigravity CLI 1.2.143 minhigheffort