Musk: Grok 4.8 enters reinforcement learning phase, limited rollout this week

Cointime update: Elon Musk said on November 10 that xAI’s Grok 4.8 has finished pretraining and entered the reinforcement learning stage, with evaluation metrics beginning to improve. He said the team abandoned JAX and SGLang and rewrote the entire RL and inference stack from scratch in C/C++, which caused delays. A limited gray-scale rollout to Bot users is planned for later this week, and Musk said Grok 4.8 will be markedly different from 4.7.