xAI completes training for 1.5T Grok V9-Medium

At 1.5T parameters, V9-Medium is a massive leap for xAI, tripling the size of their current production model to challenge the top-tier LLM hierarchy.

• –Scale represents a 3x increase over the 0.5T v8-small model, signaling a significant compute investment
• –Heavy inclusion of Cursor data during training suggests a primary focus on unseating Claude 3.5/4 as the developer's choice for complex coding
• –Optimization for NVIDIA Blackwell architecture indicates xAI is leveraging the latest hardware to manage the inference costs of such a large model
• –Supervised fine-tuning and reinforcement learning are starting immediately to meet the aggressive 3-week launch window
• –Plans to open-source the older v8-small model by year-end keeps xAI aligned with their "open weights" marketing strategy