Zendoric
← Back to the day · October 21, 2026

Replit makes OpenAI its default provider instead of Anthropic, says its CEO, though Anthropic offers nuance

🕒 Published on Zendoric: October 21, 2026 · 00:18

✨ AI-generated · how it's made

Replit, the AI coding platform that has become a favorite among startup founders and developers, has stopped working with Anthropic, according to its CEO and co-founder, Amjad Masad, speaking on stage at TechCrunch Disrupt.

🎧 Listen to the analysis (in Spanish) · synthetic voice

Replit, the AI coding platform that has become a favorite among startup founders and developers, has stopped working with Anthropic, according to its CEO and cofounder, Amjad Masad, speaking onstage at TechCrunch Disrupt. Masad attributed the decision to several factors, including cost. According to TechCrunch, Replit announced a partnership with OpenAI in August, but until now there had been no indication that it had stopped working with Anthropic.

“I think ChatGPT is better now,” Masad said. He explained that the company uses an internal benchmark to evaluate metrics such as performance, and that it also publishes a public benchmark. In his words, “consistently, Sol and Astra perform better at building ‘vibe-coded’ apps from scratch.” He summed up the decision as a combination of cost, partnership and performance, and said that for all those reasons they felt OpenAI was “a better partner for now.” The article does not explain what “Sol” and “Astra” are beyond that quote.

The break, however, does not appear to be complete. After the event, Masad clarified that customers can still access Anthropic models on the platform, but that OpenAI's ChatGPT is now the default option. TechCrunch notes that this change could significantly reduce Replit's spending with Anthropic, because most customers stick with the default model rather than choosing one themselves. Replit sells a cloud-based software development platform where users build and deploy applications directly from the browser.

Before the OpenAI deal, Replit relied on Anthropic's Claude models to power its platform. According to the article, that relationship helped the company grow its revenue significantly, and Masad said onstage that Replit had been, until recently, one of Anthropic's five largest customers. Speaking to TechCrunch over the weekend, he said: “Replit remains model-agnostic. Our users can choose Anthropic's Claude or other supported models… OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics and the strength of our partnership.”

Masad's statements were disputed by sources close to the relationship between Anthropic and Replit. A person familiar with the matter, who asked not to be identified, said the partnership remained in place. TechCrunch adds that there is public evidence that Replit was promoting Anthropic products until just a few weeks ago. For example, a blog post from late September cites Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent's production environment. Around the same time, a Replit product employee posted on X about using Opus 5.5, Anthropic's flagship model, and praised its capabilities. The article was updated after its original publication to include more context on Replit's partnerships with Anthropic and OpenAI, and its headline was changed.

Masad also shared figures on the company's growth. Although he did not disclose annual recurring revenue (ARR) figures, he said the enterprise segment accounted for around 2% of revenue last year, compared with roughly 15% this year in enterprise contracts, and that more than 50% of revenue now comes from the B2B business. Backstage, he told TechCrunch he expects more than half of Replit's revenue to come from enterprise and government customers in 2027.

Finally, Masad spoke about the need for strong AI safety measures. He compared agents to people, who need a suitable environment to work well: an office, food, air conditioning, a computer. Agents, he said, need “a habitat” with good orchestration, good error recovery, security and isolation (sandboxing), good networking, and proper identity and authorization. Once all of that is in place, he argued, AI can be deployed safely.

In short, there are two conflicting accounts. Replit presents the change as a decision based on cost, performance and partnership, with OpenAI as the default option and Anthropic models still available. Sources close to the relationship between Anthropic and Replit maintain that the partnership is still in place. The article provides no spending figures or contractual details that would resolve the discrepancy.

🔗 Related on Zendoric

Sources & references