Musk says Grok 4.8 pre-training is done, with a soft rollout to Bot users later this week
Briefing by News4U · First published
Today Elon Musk said on X that xAI has dumped JAX and SGLang and rewritten its entire reinforcement-learning and inference stack from scratch in C and C++, that Grok 4.8 pre-training is done, and that a soft rollout to Bot users is planned later this week. Musk said the rewrite took extra time, that cleanup was needed on the supervised fine-tuning data, and that 4.8 is now in reinforcement learning, with evaluation scores starting to climb. He added that Grok 4.8 will bear no resemblance to 4.7 and is not yet incredible, but that the foundations are built right and he expects it to evolve into something special. [1]