Musk has revealed that the training process for Grok's foundational model, V9-Medium (1.5T), has reached completion, boasting promising evaluation outcomes. A significant volume of Cursor data has already been integrated into the supplementary training phase, with plans to incorporate additional data in the near future. Presently, the model is undergoing fine-tuning, and reinforcement learning is slated to begin in the coming days. An official launch is anticipated within the next 2 to 3 weeks.