This is the high-reasoning-effort variant of xAI's Grok 4.5 model, jointly trained with Cursor, run inside Cursor. It is tuned to solve tasks in fewer steps and tokens, and is used for code generation, debugging, and complex multi-step agentic software engineering work.
Last 12 weeks · weekly buckets. New tools get more accurate as observation weeks add up.
Where builders who adopted this model actually run it — distribution by adopters.