Musk Calls Claude Opus 5.5 a "Masterpiece," Concedes Grok 4.8 Won't Beat It

At a glance

Elon Musk said xAI's next AI model, Grok 4.8, will likely not perform better than Anthropic's Claude Opus 5.5. He called Claude Opus 5.5 a "masterpiece." Grok 4.8 will first be released to some bot users this week.

Key points

1. 1

   Grok 4.8 has finished pretraining and is now in reinforcement learning, a process where AI improves through trial and error.

2. 2

   xAI dropped the JAX and SGLang tools and rebuilt its inference system, which generates AI answers, in C and C++.

3. 3

   Musk expects the follow-up model, Grok 5, to surpass Claude Opus 5.5 in two to three months.

4. 4

   Claude Opus 5.5 scored 66.4% on the Terminal-Bench 4.0 test, while OpenAI's GPT-6 Astra scored 57.9%.

Why it matters

This news shows how AI models compete and where the technology is heading. However, Grok 4.8's official performance scores have not been released yet.