Elon Musk doesn’t often hand out compliments to his competitors. However, on November 10 on He is now banking on Grok 5 to regain the advantage, with a target set at two or three months.
In brief
- Elon Musk recognizes that Grok 4.8 does not yet compete with Claude Opus 5.5.
- xAI has completed pre-training Grok 4.8, but is continuing reinforcement learning ahead of a limited rollout.
- Musk hopes that Grok 5 will surpass Anthropic’s model within two or three months.
Musk recognizes Anthropic’s superiority bluntly
Anthropic is preparing its IPO, but the company can also look forward to another compliment. On November 10 at 10:22 UTC, Elon Musk responded to an X user who asked him to compare Grok 4.8 to Claude Opus 5.5.
His response has the merit of frankness: “ Not as good. Opus 5.5 is a masterpiece. But Grok 5 is expected to surpass Opus 5.5, and that will happen perhaps in two or three months, with continued acceleration thereafter. »
Difficult to be more direct. The boss of xAI recognizes the delay of his model while promising a much more ambitious next generation.
The two models do not yet play in the same category. Grok 4.8 continues its development, while Opus 5.5 has already been available for several weeks. For his part, Anthropic highlights its performance in agentic programmingwith an announced cost 40% lower than that of Opus 5.
Grok 4.8, the final settings before launch
Musk also gave some details on the progress of Grok 4.8. Its pre-training is complete, but xAI is still continuing reinforcement learning. This step makes it possible to refine the model’s responses using a reward system. The team also reviewed supervision data to improve its ability to follow instructions.
Another important change: xAI abandoned JAX and SGLang in favor of an infrastructure developed mainly in C and C++. The company is thus seeking to better control the performance of its training and inference systems.
This technical choice made the developers react. When one of them said “C wins again”, Musk made it clear that it was “C with a little C++”. A nuance that obviously matters a lot behind the scenes of xAI.
On the other hand, this approach involves more internal maintenance. Discontinued tools benefit from already well-established communities and ecosystem. Musk, however, did not provide any figures to measure the expected performance gains. He simply praised the work of Starlink engineers involved in the development of this infrastructure.
Grok 5 will already have to chase new competition
The timetable announced by Musk has a small problem: Anthropic does not intend to wait for xAI to catch up.
On November 9, the company revealed Claude Fable 5.2 and Claude Mythos 5.2presented as its most successful models, particularly in programming, finance and science.
In other words, Grok 4.8 will still have to progress to compete with Opus 5.5, while Grok 5 will probably have to face an already more recent generation from its competitor.
The financial context adds another dimension to this rivalry. Anthropic filed its IPO on October 26 and launched its investor tour on November 10. The race for AI and capital is therefore accelerating on both fronts.
For Musk, the next meeting will be that of Grok 5. It remains to be seen whether xAI will meet its deadlines. In the artificial intelligence sector, two or three months can be enough to completely change the ranking.
Maximize your Tremplin.io experience with our ‘Read to Earn’ program! For every article you read, earn points and access exclusive rewards. Sign up now and start earning benefits.
