Musk: Grok 4.8 enters the reinforcement learning stage

Elon Musk stated on social media that Grok 4.8 has entered the reinforcement learning phase and plans to roll out a limited gray release to Bot users later this week.

Grok 4.8 has rewritten the tech stack using C/C++, and training and supervised fine-tuning have been completed, with evaluation metrics beginning to improve.