Replit Shifts from Anthropic, Citing OpenAI’s Superiority and Cost Efficiency

Replit, the popular AI coding platform favored by startup founders and developers, has ceased its collaboration with Anthropic, as announced by CEO and co-founder Amjad Masad on Thursday during TechCrunch Disrupt.

Masad attributed this decision to several factors, including financial considerations. While Replit revealed a partnership with OpenAI in August, there had been no prior indication of a discontinuation with Anthropic.

“I believe ChatGPT is now superior,” Masad stated, explaining that the company utilizes both internal and public benchmarks to evaluate performance. “Our public benchmark consistently shows Sol and Astra performing better for developing ‘vibe-coded applications’ from the ground up. The decision was influenced by cost, the nature of the partnership, and performance. For all these reasons, we concluded they were a better partner for the time being.”

Replit provides a cloud-based software development platform that allows users to create and deploy applications directly within their web browser. During its partnership with Anthropic, Replit leveraged the AI lab’s Claude models to power its platform, a collaboration that reportedly contributed to substantial revenue growth for the company.

However, individuals knowledgeable about the Anthropic-Replit partnership contested Masad’s assertions. One anonymous source familiar with the situation claimed the partnership remained active.

Evidence from Replit’s website, blog posts, and social media suggests the company was still employing Anthropic products just weeks ago. For example, a blog entry from late September mentioned Anthropic’s Claude Fable 5 and Claude Fable 5.1 running in Replit Agent’s production environment. Concurrently, a Replit product employee posted on X about using and promoting the capabilities of Anthropic’s flagship model, Opus 5.5.

TechCrunch has contacted Replit again for clarification and will update this article upon receiving a response.

var playerInstance_jwplayer_6ad3ba9356369 = jwplayer( “jwplayer_6ad3ba9356369” );
playerInstance_jwplayer_6ad3ba9356369.setup({
playlist: “https://cdn.jwplayer.com/v2/media/pzeTfQED”,
});

Masad also provided details on the company’s expansion during his appearance at Disrupt, highlighting that enterprise clients are now contributing a larger portion of revenue.

“We aren’t disclosing annual recurring revenue figures just yet, but regarding growth, enterprise accounted for roughly 2% of revenue last year,” Masad stated. “This year, enterprise contracts represent about 15% of our revenue, and over 50% of our total revenue is from B2B.”

Masad further emphasized the crucial role of effective AI safety measures.

He elaborated, “Just as humans require a specific environment to perform well—a comfortable office, perhaps good food, air-conditioning, a computer—agents also need a suitable habitat to exist and operate effectively. This habitat involves robust orchestration, efficient error recovery, strong safety and security through sandboxing, reliable networking, and proper identity and authorization. Once these components are in place, AI can be deployed with confidence.”

This article was updated to include additional context regarding Replit’s use of Anthropic’s models.

Leave a Reply