xAI has released Grok 4.6, an update designed for long-running agentic tasks and complex, multi-step work. This model handles research, codebase management, and application development with improved focus and accuracy. By utilizing a longer supplemental training run, the team refined how the model processes reasoning and technical data, leading to stronger performance in SFT and RL stages.
Grok 4.6 demonstrates its range by taking broad product concepts and turning them into functional versions. It can research new domains and structure applications independently, showing a tendency to verify its own work before proceeding to the next step. Visual and interactive projects also benefit from this update, as the model generates more substantial first passes that provide a solid foundation for iteration.
Safety remains a priority alongside these capabilities. The model includes an updated safety stack calibrated to match its increased intelligence, allowing for use in areas like vulnerability patching and engineering design. This includes extensive pre-deployment testing and third-party validation to maintain secure operation.
Benchmarks show Grok 4.6 matching top-tier models in the industry on the Artificial Intelligence Index. It is now available through several platforms, including Cursor and Grok Build, as well as via API for developers. To mark the launch, the company is providing double the usage allowance for the first week in Cursor and Grok Build.
For those looking to integrate the model, API documentation and access keys are available through the xAI console. Pricing is set at $2 per million input tokens and $6 per million output tokens, with a faster variant available at twice the price.