August: Replit announced a partnership with OpenAI, without signalling that its long-running reliance on Anthropic’s Claude models was ending. Claude had helped power the browser-based coding platform during a period of sharp growth, and Masad said Replit had until recently ranked among Anthropic’s five largest customers.¹

Late September: The evidence still pointed to an active Anthropic role. A Replit blog post identified Claude Fable 5 and Claude Fable 5.1 as models used in Replit Agent’s production environment, while a product employee promoted Anthropic’s Opus 5.5 on X. That trail matters because default settings, rather than merely available options, can determine where the bulk of a platform’s model spending goes.¹

Thursday at TechCrunch Disrupt: Masad said Replit was “no longer working with Anthropic,” framing the choice as a cost, partnership and performance decision. “I think ChatGPT is better now,” he said, arguing that Replit’s internal and public benchmarks showed its Sol and Astra products performing better at building vibe-coded apps from scratch.¹

The claim drew immediate pushback from a person familiar with the relationship, who said the Anthropic partnership remained in place. The disagreement is less about whether OpenAI has gained ground than about what “no longer working” actually means: a clean break, or a change in preferred supplier.¹

After the event: Masad narrowed the message. Customers can still choose Claude and other supported models, he said, but OpenAI’s ChatGPT is now the default for Replit Agent. “Replit remains model agnostic,” he said, while describing OpenAI as the preferred partner based on “performance, economics, and the strength of our partnership.”¹

That leaves Anthropic present on Replit, but no longer in the driver’s seat—a commercially significant distinction as the company targets more enterprise and government business.