Billionaire entrepreneur Elon Musk has publicly conceded that he was wrong about Dario Amodei’s Anthropic, one of the biggest rivals of SpaceXAI, a year ago.
Writing on social media as a reply to a post, Musk publicly admitted he was “clearly wrong about Anthropic.”
Musk’s humble admission came hours after SpaceXAI released Grok 4.5 to the public, acing synthetic benchmarks as far as coding is concerned.
Replying to a discussion initiated by a user regarding the competitive landscape, Musk walked back years of harsh criticism against Anthropic, which is currently leading the AI market with its Claude Fable and Opus 4.8 models.
“I was clearly wrong about Anthropic,” Musk wrote.
“They are obviously currently the leader in AI.
No company has released a model as good as Mythos/Fable and they will undoubtedly have Mythos 2 ready soon.
And I would never cut them off in a way that hurt them badly, even as a competitor.
That’s not my style,” he assured in his post.
Musk’s latest statements represent a complete 180-degree turn from his previous rhetoric.
In September 2025, he bluntly asserted that “winning was never in the set of possible outcomes for Anthropic.” More recently, in February 2026, after Anthropic raised a staggering $30 billion funding round, Musk slammed the company’s Claude models as biased and predicted that the firm was “doomed.”
Musk’s direct challenge to Anthropic followed by high praise
Over the past year, Anthropic set the tone for the commercial AI industry by tuning its AI model for the developer community, and focusing on productivity tools.
Claude, in its top-tier avatars, has been highly capable at coding, and its Fable 5 and Mythos 5 models have showcased what these models can do for cybersecurity and code development.
To make its AI product more competitive, SpaceXAI is pushing Grok into the coding space and with Cursor – the AI startup focused on vibe coding – onboard the SpaceXAI team, Musk’s team released Grok 4.5.
The latest Grok 4.5 is heavily marketed as an “Opus-class model, but faster, more token-efficient, and lower cost.”
Priced at $2 per million input tokens and $6 per million output tokens, Grok 4.5 is intentionally positioned to aggressively undercut Anthropic’s Claude enterprise pricing structure, all while offering similar levels of performance.
Yet, even as Musk praised his model’s real-world software engineering capabilities, he acknowledged that Anthropic’s Mythos 5 and Fable 5 models as the gold standard in terms of raw capability.
Independent evaluations back up Musk’s humility.
According to data from Artificial Analysis, while Grok 4.5 immediately established itself as the top-performing non-Anthropic model on the market, Anthropic’s offerings still hold the definitive edge at the top of global benchmarks.
Musk addresses the $40 billion compute question
Musk’s public reassurance eases all concerns related to cutting off Anthropic from its compute power, despite the rivalry between the two tech moguls.
Back in May 2026, Anthropic signed a major infrastructure agreement to lease the entire capacity of SpaceX’s Colossus 1 data center facility in Memphis.
Powered by over 220,000 Nvidia GPUs and pulling more than 300 MW of power, the deal requires Anthropic to pay SpaceX roughly $1.25 billion monthly through 2029, putting the total value of the contract at over $40 billion.
Previously, rumours had it that SpaceX might exercise a 90-day exit clause in the lease to stifle Anthropic’s compute resources.
Musk, however, shut down that speculation, pointing to a personal philosophy of competition based on principles.
“Tesla open sourced its patents and we made the Supercharger network available to all competitors, even though we could have made it a walled garden.
SpaceX launches competing satellite systems with no increase in price or use of unfair terms,” wrote Musk, addressing how his companies help the competition.
“Even my worst enemies can attack me on this platform,” wrote Musk, referring to harsh critics and his rivals posting on X, opposing Musk’s methods.
