Grok 4.6 has now been officially launched. Leveraging the foundation laid by Grok 4.5, this latest iteration significantly boosts its proficiency in managing prolonged agent tasks, intricate interactions, and visual projects. It can adeptly handle multi-step, complex tasks and has demonstrated top-tier performance across various benchmarks, achieving a score on par with GPT-5.6 Sol on the Artificial Analysis Intelligence Index. At present, the model is accessible on Cursor and Grok Build, and partner platforms can also integrate it via API. For the inaugural week, both platforms are offering double the standard usage quota. Grok 4.6 has been subjected to extensive supplementary training and undergone multi-stage refinement. In real-world testing, it showcases exceptional ability in translating broad product concepts into functional prototypes and independently carrying out testing and validation processes. Notably, it has made substantial strides in visual and interactive projects. From a safety perspective, the model has fine-tuned its protective mechanisms and undergone the most comprehensive safety evaluation to date. As for pricing, it is set at $2 per 1 million input tokens and $6 per 1 million output tokens, with prices doubling for accelerated versions.