Elon Musk publicly conceded on X that xAI's upcoming Grok 4.8 model will be "not as good" as Anthropic's Claude Opus 5.5, calling the rival system "a masterpiece." The comments came in a reply to a user named Dimdv on Tuesday.
Musk said Grok 4.8 has completed pretraining but remains in reinforcement learning, with the team also spending significant time cleaning supervised fine-tuning data. He detailed a major engineering shift: xAI dropped the JAX machine-learning framework and SGLang serving software in favor of an inference and reinforcement-learning system rewritten in C and C++, crediting Starlink engineers for contributions to the Grok training and inference stack.
He described Grok 4.8 as "not yet incredible" but said its foundations were finally built correctly, with a soft rollout to Bot users expected later this week. Musk predicted that the next-generation Grok 5 "should exceed Opus 5.5" possibly within two or three months, though he provided no firm release date or benchmark scores.
Anthropic released Claude Opus 5.5 on September 22, claiming performance comparable to its higher-end Fable 5.1 model while costing 40% less to run than the previous Opus version. On Terminal-Bench 4.0, Anthropic reported 66.4% for Opus 5.5 versus 57.9% for OpenAI's GPT-6 Astra, while Grok remains far behind on public leaderboards.
