Replit, the browser-based AI coding platform, has switched its default model provider from Anthropic to OpenAI, CEO Amjad Masad said on stage at TechCrunch Disrupt on Thursday. In follow-up comments over the weekend he clarified that Replit is not cutting Anthropic off: users can still choose Claude models, but OpenAI is now the default behind Replit Agent.

What was said

TechCrunch reports that Masad initially said Replit is “no longer working with Anthropic,” and gave several reasons. “It was a decision based on both cost, partnership, and performance,” he said, adding that Replit uses an internal benchmark and also publishes a public one, and that “consistently, Sol and Astra perform better on building vibe-coded applications from scratch.” Replit announced a partnership with OpenAI in August, although at that point there was no sign it was ending its work with Anthropic.

After Disrupt, Masad adjusted the message. Customers can still access Anthropic models on the platform, he told TechCrunch, but OpenAI’s models are the default. “Replit remains model agnostic,” he said. “OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.”

Why the default matters

The distinction between “no longer working with” and “no longer the default” is more than semantics. Most users of a coding tool never change the model setting, so the default determines where most of the usage, and therefore most of the spending, goes. TechCrunch notes the shift could significantly reduce Replit’s spend with Anthropic, which Masad said had been a top-five customer until recently. Earlier reporting described the Anthropic relationship as having helped Replit grow revenue significantly.

A disputed account

The story is contested. TechCrunch says Masad’s on-stage statements were rejected by insiders familiar with the Anthropic–Replit relationship, and one person said the partnership was still in place. The article also points to Replit’s own blog and social posts as recently as late September that cited Anthropic models running in Replit Agent’s production environment. We have no independent confirmation of the contract terms on either side, so the reliable facts are limited to what Masad said publicly and that users can still select Claude.

The business picture

Masad also shared a sense of Replit’s growth shift. He said enterprise was roughly 2% of revenue last year and about 15% this year, with more than half of revenue now coming from business customers, and that he expects more than half of revenue to come from enterprise and government in 2027. Replit has not disclosed annual recurring revenue figures. He described agents as needing “a habitat,” meaning orchestration, error recovery, safety and sandboxing, an argument for platforms like Replit that wrap models in infrastructure rather than just calling them.

Why it matters

For developers, this is a reminder that AI coding products are model-agnostic layers. The model underneath can change with a price negotiation or a benchmark result, and behavior can shift with it. Teams that depend on a specific model’s style should pin it explicitly rather than rely on a default. For model labs, it shows how much distribution partnerships matter: being the default inside a popular tool is a large revenue channel, and losing it is costly. The competition between labs for coding workloads, which is where much of the commercial demand sits today, is intensifying. Our overview of how AI coding assistants are best used and our explainer on vibe coding give background on why this market matters.

What to watch next

  • Whether Anthropic or Replit issues a formal statement resolving the disagreement over the partnership.
  • Independent results for Replit’s public benchmark on building apps from scratch.
  • Whether Replit adds or changes model options for Replit Agent.
  • Whether other coding platforms re-open their default model choice.
  • Any disclosure of Replit’s annual recurring revenue.

Key takeaways

  • OpenAI models are now the default for Replit Agent, according to CEO Amjad Masad.
  • Anthropic models remain selectable on the platform.
  • Masad cited cost, partnership and performance; performance claims rely on Replit’s own benchmark.
  • Insiders disputed the on-stage description of the Anthropic relationship.
  • Enterprise is growing quickly as a share of Replit’s revenue.

Frequently asked questions

Can Replit users still use Claude?

Yes. Masad said customers can choose Anthropic’s Claude or other supported models; the change is to the default.

Did Replit end its Anthropic partnership?

That is disputed. Masad said on stage it had ended, then described Replit as model agnostic; an unnamed person familiar with the relationship told TechCrunch it was still in place.

Why do defaults matter so much?

Because most users never change them. The default model receives the bulk of requests, which drives both quality perception and what Replit pays each provider.