Replit is no longer working with Anthropic, CEO says
TechCrunch Lucas Ropek, Kirsten Korosec
Replit’s CEO says it’s no longer working with Anthropic. He says cost and performance pushed the switch, but people familiar with the deal say Anthropic is still in play.
Based on reporting by TechCrunch, Lucas Ropek, Kirsten Korosec — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Replit’s AI coding platform may be at a turning point, but its partnership story is messier than the CEO made it sound onstage. At TechCrunch Disrupt on Thursday, co-founder and CEO Amjad Masad said the company is no longer working with Anthropic, and framed the move as a mix of cost, partnership issues, and performance. He also said, plainly, that “ChatGPT is better now.”
That is not a small shift for a company whose browser-based development platform has relied on Anthropic’s Claude models to power its service. Replit has become popular with startup founders and developers who want to build and deploy apps directly from a browser, and Masad said the Anthropic tie-up had helped the company grow revenue significantly. But he also pointed to OpenAI, announcing a partnership with that company in August, and said Replit uses an internal benchmark, plus a public one, to compare model performance.
The twist is that people familiar with the Anthropic relationship pushed back almost immediately. One person with knowledge of the arrangement, speaking anonymously, said the partnership was still in place. There’s also a paper trail: a September blog post cited Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment. So either the transition is not complete, or the public message got ahead of the operational reality.
Masad also used the stage time to show how Replit’s business is changing. He said enterprise accounted for about 2% of revenue last year and about 15% this year, while more than half of revenue is now B2B. He also argued that AI agents need the right “habitat” to work safely, with orchestration, error recovery, sandboxing, networking, identity, authorization, and other controls in place. That’s the kind of talk everyone in AI loves to give. The harder part is proving the system underneath is actually as clean as the slide deck.
My take — AI-written commentary, not fact-checked reporting
This smells like the new AI vendor shuffle: declare one model family stale, praise the next one, and hope the logs don’t spoil the story. The real tell is not which chatbot sounds best onstage, but which one is still in production when the applause dies down. Replit may be making a sensible business move; it’s also doing the usual Silicon Valley thing of calling a messy transition a strategic pivot.
Read more about this at: TechCrunch