Musk: Grok 4.8 is in reinforcement learning, with a soft rollout to Grok Bot users planned this week

Elon Musk said on X that xAI has dropped JAX and SGLang and rewritten its entire reinforcement-learning and inference stack from scratch in C/C++, which took extra time. According to Musk, pre-training of Grok 4.8 is done, the SFT data needed a lot of cleanup, and the model is now in RL with its evals "starting to climb".

Musk says xAI will do a soft rollout to Bot users later this week. He says Grok 4.8 "will bear no resemblance to 4.7" and is "not yet incredible", but that the foundations are now built right. These are Musk's statements; xAI has not published a release or benchmarks.