Replit has stopped working with Anthropic, says CEO Amjad Masad
The AI coding platform's chief told a San Francisco audience that cost and performance drove the switch, but people close to the partnership say it is still in place.
Key points
- Amjad Masad said Replit is no longer working with Anthropic, citing cost, partnership and performance.
- People familiar with the deal said the partnership was still in place, and Replit was promoting Claude models weeks ago.
- Masad said 15% of Replit's revenue now comes from enterprise contracts, up from about 2% last year.
Replit, the AI coding platform popular with startup founders, has stopped working with Anthropic, its chief executive Amjad Masad said, a claim that people close to the partnership dispute.
Masad made the remark on stage at TechCrunch Disrupt in San Francisco. He said the decision came down to several things, cost among them. “I think ChatGPT is better now,” he said.
Replit announced a partnership with OpenAI in August. Until now there had been no sign that it had broken with Anthropic, whose Claude models have powered much of its platform and helped it grow revenue quickly.
What Masad said
Masad said Replit judges models on an internal benchmark and publishes a public one too. “Consistently, Sol and Astra perform better on building vibe-coded applications from scratch,” he said. “It was a decision based on both cost, partnership, and performance. And for all these reasons, we just felt like they were a better partner for now.”
He did not give revenue figures but said the business was shifting toward companies. “Enterprise was like 2% of revenue last year,” he said. “This year, I would say like 15% of our revenue is enterprise contracts, and more than 50% of our revenue is B2B.”
A disputed break
People familiar with the arrangement rejected his account. One said the Anthropic partnership was still in place.
Replit’s own material points the same way. A Replit blog post from late September listed Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in the production environment of Replit Agent, its app-building assistant. Around the same time, a Replit product employee posted on X promoting Anthropic’s flagship model, Opus 5.5. Neither Replit nor Anthropic had said anything further by Saturday.
The episode shows how fluid the market has become for coding tools that rent their intelligence from a few large AI labs. Platforms such as Replit can switch suppliers, or play them off against each other, as prices and benchmark scores move.
Agents need a “habitat”
Masad also spoke about how Replit itself runs. He described it as a “self-driving company”, where AI agents do the first pass on work such as triaging support tickets and drafting fixes, and people step in to decide. “You need human judgment; you need taste,” he said.
He argued that agents can only be used safely inside the right setting. “Agents need a habitat for them to exist and do good work,” he said, listing orchestration, error recovery, sandboxing, networking and identity controls. “And once you figure all of that out, then you can deploy AI safely.”