Replit is no longer working with Anthropic, CEO says

Replit is no longer working with Anthropic, according to the AI coding platform's CEO and co-founder Amjad Masad, who said so Thursday onstage at TechCrunch Disrupt. Masad attributed the change to several factors, naming cost among them.
Replit had announced a partnership with OpenAI in August, and there had been no prior indication that its relationship with Anthropic had ended. The two events are not explicitly linked in Masad's remarks, but the timing suggests the OpenAI deal marked the transition rather than a simple addition of a second model supplier.
Performance was central to the explanation. Masad said, in effect, that ChatGPT is better now. He described an internal benchmark the company maintains to assess metrics such as performance, and said Replit also publishes a public benchmark. According to Masad, the models he referred to as Sol and Astra consistently do better at building vibe-coded applications from scratch. For developers, that phrasing points to a specific evaluation target: generating working applications from natural-language prompts with minimal manual coding, rather than assisting with edits inside an existing codebase. He characterized the outcome as a decision resting on cost, partnership and performance together, concluding that the other vendor felt like a better partner for now.
Anthropic's Claude models had powered Replit's platform under the partnership, and that arrangement was described as having helped Replit grow revenue significantly. The reversal is therefore notable given how closely the two companies had been associated.
Masad's account drew pushback from people familiar with the Anthropic-Replit relationship. One person, who asked not to be named, said the partnership was still in place. There is also evidence that Replit was using Anthropic products as recently as September: a blog post from that month cites Anthropic's Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent's production environment. TechCrunch said it reached back out to Replit for clarification and would update the story.
The practical effect is a statement of intent from one executive at a public event, contested by an unnamed source, and not accompanied by a joint announcement from either company. For teams tracking which model providers power which coding products, that leaves the operational picture unsettled: a production environment described in September as running Anthropic models, a CEO saying Thursday that the company no longer works with Anthropic, and a person familiar saying otherwise.
Masad also discussed Replit's growth. He declined to share ARR numbers, but offered a revenue-mix comparison: enterprise was about 2% of revenue last year, he said, while this year roughly 15% of revenue comes from enterprise contracts, with more than 50% of revenue coming from B2B overall. Those figures describe a shift toward larger paying customers and business buyers rather than individual developers, which is a common trajectory for developer tools that start bottom-up and later court procurement-driven accounts. The enterprise share is small in absolute terms but has moved several-fold in a year.
He additionally spoke about AI safety, using a habitat metaphor: just as people need a decent environment to work in, agents need a habitat to exist and do good work. He listed the components of that habitat as good orchestration, error recovery, safety and security with sandboxing, networking, and identity and authorization. Solve those, he said, and you can deploy AI safely. This framing treats safety less as a model-level property and more as infrastructure around the model, which is a useful lens for platform teams deciding what they must build themselves.
Why it matters: Replit is a widely used browser-based environment for building and deploying apps, so which models run inside Replit Agent affects the output and cost profile developers see. If the switch holds, users should expect different behavior from the agent and possibly different pricing dynamics, and any product that standardized on Claude through Replit may need to re-evaluate. The competing accounts mean buyers should treat the change as likely but not confirmed, and verify current model backing in their own projects before committing to a long-lived workflow. Enterprise and B2B revenue now make up a much larger share of Replit's business, which in my reading makes vendor reliability and procurement-friendly terms more important to the platform than they were a year ago.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Masad said the decision was made for a number of different reasons, including cost. Replit announced a partnership with OpenAI in August, but there was never any indication that it was no longer with Anthropic.