Replit makes OpenAI its default over Anthropic, CEO says

Replit, the browser-based coding platform used heavily by startup founders and developers, has shifted its default model to OpenAI's ChatGPT, CEO and co-founder Amjad Masad said Thursday onstage at TechCrunch Disrupt. Masad said the change was driven by several factors, with cost among them. Replit had announced a partnership with OpenAI in August, but until his remarks there was no public sign that its work with Anthropic had changed.
Masad was direct about the reasoning. He said ChatGPT is better now, and that Replit assesses performance with both an internal benchmark and a public one. On building what he called vibe-coded applications from scratch, he said the models Sol and Astra consistently perform better. He described the outcome as a decision resting on three legs: cost, partnership and performance, adding that the team felt OpenAI was a better partner for now.
Anthropic has not been fully removed from the product. After his appearance, Masad told TechCrunch that customers can still get to Anthropic models on the platform, but OpenAI's ChatGPT is now the default. That is a meaningful distinction for Anthropic's revenue. Most users of a developer platform tend to accept whatever model is selected for them rather than switching manually, so changing the default can move a large share of usage, and therefore spend, from one provider to another. Before the OpenAI deal, Replit ran its platform on Anthropic's Claude models, a relationship described as having helped the company grow revenue substantially. Masad said onstage that until recently Replit had been a top-five Anthropic customer.
He framed the arrangement as one of preference rather than exclusivity, telling TechCrunch that Replit remains model agnostic and that users can pick Claude or other supported models, while OpenAI is now the preferred partner for powering Replit Agent on the basis of performance, economics and partnership strength.
That account was contested. People familiar with the Anthropic-Replit relationship rejected Masad's onstage characterization, and one person, speaking on condition of anonymity, said the partnership was still in place. There is also public evidence that Replit kept promoting Anthropic products close to the announcement: a late-September blog post cited Claude Fable 5 and Fable 5.1 as models running in Replit Agent's production environment, and around the same time a Replit product employee posted on X about using Anthropic's flagship Opus 5.5 and its capabilities. Those details point to a gradual repositioning rather than a clean break, and readers should treat the two accounts as unresolved.
Masad also gave a rare look at how Replit's revenue is composed. He declined to share ARR figures, but said enterprise was about 2% of revenue last year and around 15% this year in enterprise contracts, with more than half of revenue now coming from B2B customers. Backstage, he told TechCrunch he expects enterprise and government customers to supply more than half of revenue in 2027. For a company that began as a tool for individual developers and small teams, that is a notable tilt toward larger contracts, which typically bring longer sales cycles but steadier, more predictable spending.
On AI safety, Masad used a habitat analogy: just as people need an office, food, air-conditioning and a computer to work well, agents need an environment built for them. In his description that habitat includes good orchestration, error recovery, safety and security with sandboxing, networking, and identity and authorization. Once those pieces are in place, he argued, AI can be deployed safely. This suggests Replit is pitching its platform layer, not just model access, as the differentiator, since the model itself is increasingly interchangeable.
In practice, the model-default decision is the sort of choice more platform companies are likely to face as leading models converge in quality and pricing pressure builds. Making a default is a strong signal to users, and it can reshape a vendor's revenue without any formal termination of a partnership.
The article was originally published on October 16 at 2:22 p.m. PT and was later updated to add context on the Anthropic and OpenAI relationships and to change its title.
Why it matters: Replit's users who rely on the default will now be routed to OpenAI models unless they deliberately choose otherwise, which changes both the out-of-the-box experience and which vendor captures most of that usage spend. The disputed accounts of the Anthropic relationship suggest developers should verify model availability in their own projects rather than assume a partnership has ended. For product teams, the enterprise and B2B revenue shift Masad described hints that Replit is optimizing for organizational buyers, and it is a reasonable inference that model routing, safety and orchestration will be sold as part of that pitch.
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 partnering with Anthropic.