Elon Musk said Grok 4.8 has finished pretraining, is now in reinforcement learning, and will get a limited rollout to Grok Bot users later this week, Tesla North reported, quoting his Nov. 10 posts on X. TokenPost reported the same comments.
Why it matters: Musk said xAI had “dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++,” work that “took some extra time,” and that a lot of cleanup was needed on the supervised fine-tuning data. He did not announce a wider public release date, and xAI has not posted its own release notes for the model.
“Now, 4.8 is in RL and the evals are starting to climb,” Musk wrote, according to both outlets. He later called the new stack “C with a little C++, but not too much,” said it includes some assembly, and described it as “tight code.” “Grok 4.8 will bear no resemblance to 4.7 (thankfully),” he wrote. The new version is “not yet incredible,” he added, but “the foundations are built right.”
Asked how 4.8 would compare with Anthropic’s Opus 5.5, Musk said it was “not as good” and called Opus 5.5 a “masterpiece,” Tesla North reported. He said Grok 5 “should exceed Opus 5.5 and maybe that happens in 2 or 3 months.” In mid-September, Musk had described 4.8 as a 2.5 trillion-parameter model that would finish training that week, a timeline that has since slipped.