Elon Musk has praised a rival model while conceding his own next release is not ready to match it. Asked on X how Grok 4.8 would compare with Anthropic’s Claude Opus 5.5, the xAI founder said it would be not as good. In the same exchange, he called Opus 5.5 a masterpiece. The reply came Tuesday to a user named Dimdv.
Grok 4.8 is still in training
Musk said Grok 4.8 had finished pretraining but was still going through reinforcement learning. That phase uses feedback and rewards to refine a model’s answers. He added that the team spent a lot of time cleaning supervised fine-tuning data, the examples used to teach a model how to follow instructions.
He also said xAI dropped JAX and SGLang, a machine-learning tool and model-serving software, in favor of an inference and reinforcement-learning system rewritten in C and C++. Inference is the step that generates an answer after a user sends a prompt. Rewriting that software can give engineers more control, though Musk did not share numbers showing the gain.
A developer named tetsuo cheered the news with C wins again. Musk replied that the code was C with a little C++. He later credited Starlink engineers with work that helped build the Grok training and inference stack. He pointed to SpaceX’s habit of writing its own headers and handling precise time calculations, including leap seconds.
Musk described Grok 4.8 as not yet incredible. Still, he said its foundations are finally built correctly. He expects a soft rollout to Bot users later this week. That would give xAI an early test before any wider release.
Anthropic has a lead
Anthropic released Opus 5.5 on September 22. The company says it performs close to its higher-end Fable 5.1 model on most work while costing 40% less to run than the previous Opus version on typical workloads. It is aimed at difficult programming jobs and tasks where AI agents work through multiple steps with limited human help.
On Terminal-Bench 4.0, which checks whether AI agents can finish complex command-line tasks, Anthropic reported 66.4% for Opus 5.5. OpenAI’s GPT-6 Astra scored 57.9%. Grok is far behind on the leaderboards.
Musk has not published Grok 4.8 scores that would show the size of the gap. His near-term target is the limited rollout this week. He says Grok 5 should exceed Opus 5.5, perhaps in two or three months. He has not given a firm date, and he has a history of missing his own targets.
