Replit CEO says company has stopped working with Anthropic; insiders dispute it

Replit CEO and co-founder Amjad Masad said Thursday at TechCrunch Disrupt that the AI coding platform is no longer working with Anthropic. He said the decision rested on several factors, including cost, partnership and performance. Replit announced a partnership with OpenAI in August, but at the time there was no indication it had parted ways with Anthropic.

Masad said Replit uses an internal benchmark to assess performance and also publishes a public one. By his account, OpenAI's models, which he called Sol and Astra, consistently do better at building vibe-coded applications from scratch. Under the earlier arrangement, Replit used Anthropic's Claude models to run its platform.

TechCrunch reported that insiders familiar with the Anthropic-Replit relationship rejected Masad's account. One person, who asked not to be named, said the partnership was still in place. The outlet also found that a late-September Replit blog post cited Claude Fable 5 and Claude Fable 5.1 as running in Replit Agent's production environment, and that a Replit product employee had recently promoted Anthropic's Opus 5.5 on X.

In a separate session, Masad gave a look at Replit's business. He said enterprise made up about 2% of revenue last year and about 15% this year, and that more than 50% of revenue is B2B. He declined to share ARR figures.

Model suppliers compete hard for coding platforms, which generate heavy, recurring usage and visible showcases for model quality. A public statement from a prominent customer that one lab outperforms another on cost and quality is the kind of claim both OpenAI and Anthropic care about, which makes the conflicting accounts notable.

The discrepancy is unresolved. Masad's statement is categorical, yet the reporting shows Anthropic models in production weeks ago and an insider says the partnership continues. One possible reading, which the coverage does not confirm, is that the transition is under way or that the scope of the change is narrower than described.

Masad also used the stage to describe what he calls a 'self-driving company,' where agents handle first-pass work such as support triage using GitHub issues, the codebase and Linear tickets, and humans decide what to do. He framed this as shifting people toward architecture and deciding what to work on, rather than replacing them.

TechCrunch says it will update its story if Replit responds to its request for clarification. A clear statement on which models power Replit Agent, and whether any Anthropic contract remains, would settle the question.

Also worth watching is whether Anthropic comments, and whether Replit's public benchmark results are updated to show the comparison Masad described.