The Grok and Claude logos (illustrative). Image: Grok logo via Wikimedia Commons (public domain); Claude logo: Anthropic
Elon Musk has conceded that Grok is behind Claude. In an interview with China Media Group (CMG) at Tesla’s engineering headquarters, he said SpaceXAI’s latest model, Grok 4.7, is “not as good as” Anthropic’s new Opus 5.5, and that he expects to catch up “sometime next year,” according to a transcript of the interview.
“A solid workhorse of a model”
Asked about Grok 4.7, Musk was unusually modest:
It’s called Grok 4.7, a solid workhorse of a model. It’s not as good as, say, Opus 5.5 that just got released. But also, we’ve only been around for about 3 years doing AI. Whereas Anthropic’s been around for 6 years.
Elon Musk, CEO of SpaceXAI
(Anthropic was founded in 2021, so it’s closer to five.) Musk added that SpaceXAI releases a new model “roughly every month or 2,” and that it would “probably catch up to Frontier sometime next year, most likely.” Opus 5.5 came out on September 22 and has been widely praised, while Grok 4.7 has been pitched mainly on price.
Grok Bot “growing about 100% a month”
Musk was more upbeat about Grok Bot, SpaceXAI’s AI agent, which he called “kind of like a personal digital assistant.” He said it “has been very popular” and is “growing about 100% a month.” SpaceXAI hasn’t published user numbers, so that figure can’t be checked. The company has been pushing hard for users, putting Grok Bot into Tesla cars, paying creators when people use their bot templates, and offering a trip to a Starship launch to users who share theirs.
A bet on “real-world engineering”
Musk’s plan to close the gap is data only his companies have. He argued that Anthropic’s strength is code, and that SpaceX and Tesla could give Grok an edge in the physical world:
What Anthropic did extremely well was make AI excellent at software engineering, but no one has yet made AI excellent at real-world engineering.
Elon Musk, CEO of SpaceXAI
China “by far the best” for its compute
Musk also praised Chinese AI labs, saying their models are “generally outstanding” and that China is “doing by far the best in terms of performance per unit of compute.” He predicted China will solve its chipmaking limits “in about 2 or 3 years,” much faster than US export controls assume.
On safety, he suggested a US-China “working committee,” because rules in one country alone won’t work: “It’s got to be, how do we apply regulation fairly to AI produced in any country?” He put the chance of a good outcome from AI at about 90%, but said “we don’t want to get complacent regarding the 10% probability of a bad outcome.”
Why it matters
It’s rare for Musk to admit a rival is ahead, and Anthropic is the one he chose to name. With SpaceXAI adding hundreds of thousands of GPUs to Colossus 2, “next year” is the target to watch, and a clear sign that Claude, not ChatGPT, is now the model the others measure themselves against.
Sources: CMG interview transcript via The Singju Post, Wccftech.
