Replit CEO Amjad Masad said on October 16, 2026, that Replit has ended its partnership with Anthropic and now relies on OpenAI's models instead, citing cost, the terms of the partnership, and performance on Replit's own benchmarks. A person familiar with the matter told TechCrunch the Anthropic partnership is still in place, and TechCrunch found evidence that Replit was using Anthropic's models in production as recently as late September.
Masad Says Cost and Performance Ended the Replit Anthropic Partnership
Masad made the statement onstage at TechCrunch Disrupt 2026 in San Francisco. He said Replit's internal and public benchmarks show OpenAI's models perform better than Anthropic's at building applications from scratch through so-called vibe coding. He called OpenAI "a better partner for now." Replit announced a partnership with OpenAI in August 2026, and Masad gave no indication beforehand that the Anthropic relationship was ending.
A Source Disputes the Claim, and Recent Posts Show Anthropic Models Still in Use
An unnamed person familiar with the Replit-Anthropic partnership told TechCrunch it remains in place, directly contradicting Masad's onstage statement. TechCrunch also pointed to a late-September Replit blog post that named Claude Fable 5 and Claude Fable 5.1 as running in Replit Agent's production environment, and to an X post from a Replit product employee promoting Anthropic's Opus 5.5 around the same time. TechCrunch said it had contacted Replit for clarification and would update its story if the company responded.
Replit's Enterprise and B2B Revenue Have Grown Since the Anthropic Partnership
Masad gave figures without disclosing Replit's overall revenue. Enterprise contracts made up about 2% of Replit's revenue last year and about 15% this year, he said, and more than half of Replit's revenue now comes from business customers rather than individuals. A separate report from Inc. has credited Replit's partnership with Anthropic with contributing to that revenue growth.
Masad Says AI Agents Need a Safety "Habitat" Before Wider Deployment
Masad also said, separately from the Anthropic dispute, that AI agents need a supportive operating environment before they can be deployed safely at scale. He listed orchestration, error recovery, security, sandboxing, networking, and identity and authorization controls as the pieces that environment needs.
Why This Matters for Engineering Leadership
Masad's statement and the pushback it drew illustrate a problem engineering leaders face when a platform they depend on changes its model vendor: the public announcement and the technical reality can diverge, at least for a time. Teams building on Replit, or evaluating any AI coding platform's model backend, have reason to ask directly which models run in production today rather than relying on a vendor's public statement about its partnerships. Masad's list of what an agent "habitat" requires, orchestration, error recovery, sandboxing, and identity controls, is also a reasonable checklist for any team weighing whether to hand more autonomy to coding agents in its own pipeline.
FAQ
Did Replit end its partnership with Anthropic?
Replit CEO Amjad Masad said onstage at TechCrunch Disrupt that Replit is no longer working with Anthropic, but a person familiar with the matter told TechCrunch the partnership is still in place.
Why did Masad say Replit moved away from Anthropic?
He cited cost, the terms of the partnership, and performance, saying Replit's benchmarks show OpenAI's models build applications from scratch better than Anthropic's.
Is there evidence Replit still uses Anthropic's models?
Yes. TechCrunch found a late-September Replit blog post describing Claude Fable 5 and 5.1 running in Replit Agent's production environment, and a Replit employee's X post promoting Anthropic's Opus 5.5.
When did Replit partner with OpenAI?
Replit announced its partnership with OpenAI in August 2026.
How has Replit's revenue mix changed?
Masad said enterprise contracts grew from about 2% of revenue last year to about 15% this year, with more than half of revenue now coming from business customers.
What does Masad say AI agents need before wider deployment?
He said agents need a supportive "habitat," including orchestration, error recovery, security, sandboxing, and identity and authorization controls.