Elon Musk publicly called Anthropic's Claude Opus 5.5 a "masterpiece" — a rare acknowledgment of a competitor's superiority. At the same time, he confirmed that xAI's own Grok 4.8 model is inferior to Anthropic's new release.
The statement came on Tuesday in response to an X user's question about how Grok 4.8 compares to Anthropic's model. Musk's reply was extremely concise: "Opus 5.5 is a masterpiece," with the clarification that Grok 4.8 is "worse."
What's behind the Opus 5.5 release
Anthropic released Opus 5.5 on September 22. The company claims that for most tasks the model is on par with the more expensive Fable 5.1 version. The key argument is operating economics: the cost of running Opus 5.5 under typical workloads is 40% lower compared to the previous Opus version.
On the Terminal-Bench 4.0 test, Anthropic achieved a result of 65.76%. For comparison: OpenAI's GPT-6 Astra scored 57.9% on the same test. Notably, xAI has not published any benchmarks for Grok 4.8 — the absence of measurable data is itself a telling signal.
Betting on Grok 5
Musk stated that Grok 5 "should surpass Opus 5.5," possibly within two to three months, though he did not give a specific release date. Historically, the head of xAI has not always met his own deadlines, so such promises should be treated with caution.
xAI is currently finalizing work on Grok 4.8. According to Musk, the pre-training phase is complete and the model is in the reinforcement learning phase. The team paid special attention to data cleaning for supervised learning, and also abandoned JAX and SGLang in favor of a new system built in C and C++.
Musk himself characterized Grok 4.8 as "not outstanding so far" as an AI, while noting that the model has a solid foundation. The businessman expects a test launch for early users as soon as this week. Anthropic has not publicly commented on his statements so far.
Cryptalist analysis: Musk's public acknowledgment is not merely an emotional gesture but an indicator of intensifying competition in the frontier model segment, where inference economics (a 40% cost reduction) is becoming no less important a factor than raw benchmarks. For the market, this is a signal that AI leadership is shifting from marketing claims to measurable efficiency metrics, and that promises of future releases without confirmed data will increasingly rarely be perceived by investors as grounds for revaluation.