Replit says cost and performance drove its move away from Anthropic

Replit CEO Amjad Masad says the AI coding platform is no longer working with Anthropic, citing cost, partnership terms and model performance. He said Replit’s internal tests found OpenAI’s Sol and Astra models stronger for building complete applications from prompts, following a partnership with OpenAI announced in August.

The account is contested. People familiar with the Replit-Anthropic relationship told TechCrunch that the partnership remains in place. Replit’s own materials also referred to Claude Fable 5, Fable 5.1 and Opus 5.5 in production or promotional posts as recently as late September. TechCrunch asked the company for clarification.

Replit provides a browser-based environment where users can create and deploy software with AI assistance. Its earlier use of Claude was widely associated with the platform’s revenue growth, so a full supplier change would be a notable shift in the market for coding models. Masad said enterprise contracts now represent about 15 percent of Replit’s revenue, up from roughly 2 percent last year, while business customers account for more than half.

For now, the firm’s current model mix is not fully resolved. The public CEO statement supports a move toward OpenAI, but the conflicting evidence makes it premature to describe Anthropic as definitively removed from every Replit workflow.