Elon Musk has given a rival AI model an unusually enthusiastic endorsement while conceding that his own company's next release will not be able to beat it.

Asked on X how Grok 4.8 would compare with Anthropic's Claude Opus 5.5, the xAI founder replied that it would be "not as good."

"Opus 5.5 is a masterpiece," Musk said in a post on X.

The admission came in a reply on Tuesday to a question from a user named Dimdv. Musk added that Grok 5, the next-generation model set to follow this update, "should exceed Opus 5.5," potentially within two or three months. He offered no further details to support that prediction and no firm release date—a notable caveat given his well-documented record of missing his own target timelines.

The distinction matters for a company trying to catch up with Anthropic and OpenAI. By his own framing, Opus 5.5 now serves as the public yardstick for xAI's next two releases—the shipped Anthropic model that Grok 4.8 will fall short of, and the bar Grok 5 "should exceed." Grok 4.8 is the update Musk expects to place in some users' hands this week; Grok 5 is the model he believes could eventually put xAI ahead—until a competing AI lab ships a newer model.

Why Grok 4.8 Is Taking Longer

The exchange began with a technical update in which Musk said Grok 4.8 had finished pretraining, the stage in which a model absorbs knowledge from large volumes of training data, but was still undergoing reinforcement learning (RL), the phase in which a model's responses are refined using feedback and rewards. He said the team had also spent considerable time cleaning up the supervised fine-tuning data—the examples used to teach the model how to follow instructions.

Musk said xAI had abandoned JAX, a machine-learning framework, and SGLang, software used to serve AI models, in favor of an inference and reinforcement-learning system rewritten in C and C++. Inference is the process of generating answers after a user submits a prompt; rewriting the underlying software can give engineers more control over performance, though Musk provided no figures demonstrating the resulting gains.

Developer tetsuo celebrated the announcement with "C wins again," prompting Musk to clarify that the code was "C with a little C++." He subsequently credited Starlink engineers with work that helped produce the Grok training and inference stack, pointing to SpaceX's practice of writing its own headers and handling precise time calculations, including leap seconds.

Musk described Grok 4.8 as "not yet incredible" but said its foundations were finally built correctly. He expects a soft rollout to bot users later this week, giving the company an early test before any broader release.

Anthropic's Head Start

Anthropic released Opus 5.5 on September 22, delivering performance comparable to its higher-end Fable 5.1 model on most workloads while costing 40% less to run than the previous Opus version on typical work. The model targets demanding programming jobs and tasks in which AI agents work through multiple steps with limited human assistance.

On Terminal-Bench 4.0, which tests whether AI agents can complete complicated tasks in a command-line environment and scores them by the percentage successfully finished, Anthropic reported 66.4% for Opus 5.5 versus 57.9% for OpenAI's GPT-6 Astra. Grok sits far behind on the leaderboards.

Musk has not published Grok 4.8 scores that would establish how large the gap actually is. Any benchmark figures xAI discloses around this week's rollout would therefore provide the first concrete basis for measuring that gap against the results Anthropic has already made public. His immediate target is the limited rollout later this week; the promised Grok 5 challenge to Opus 5.5 remains, by his own estimate, two or three months away.