New page: Grok 4.6 — frontier model from Cursor and SpaceXAI for complex coding and knowledge work, with improved instruction following and long-horizon agentic work vs Grok 4.5; effort levels xhigh, high (default), medium, and low; stan…
Changelog
Models
New page: Grok 4.6 — frontier model from Cursor and SpaceXAI for complex coding and knowledge work, with improved instruction following and long-horizon agentic work vs Grok 4.5; effort levels xhigh, high (default), medium, and low; standard on-demand pricing $2/$0.50/$6 per M tokens and Fast $4/$1/$12; 50% launch discount for one week starting August 12, 2026; included in the Cursor Models pool.
Grok 4.5 — docs now recommend Grok 4.6; Grok 4.5 remains available; Fast output pricing documented at $12/M tokens (was $18/M); pool membership updated to include Grok 4.6.
Composer 2.5, Models & Pricing, and Teams pricing — Cursor Models pool and first-party / Cursor Token Rate exemptions updated to include Grok 4.6 alongside Grok 4.5 and Composer 2.5; Start plan notes cover all three models at non-fast / fixed medium effort for Grok.
Evals & SDK
Evals — default eval model is now Grok 4.6 (grok-4.6) in examples and guidance.
Admin API
Organization Admin API and Teams Admin API — cursorTokenFee docs list Grok 4.6 with other first-party Cursor models that do not include the Cursor Token Rate.
