Replit has made OpenAI its preferred provider and default choice for Replit Agent, while continuing to offer Anthropic’s Claude models to customers who select them. The clarification narrows an earlier statement from chief executive Amjad Masad that suggested the companies were no longer working together.
Masad said Replit chose OpenAI based on cost, partnership terms and performance in the company’s internal and public evaluations. He argued that OpenAI models performed better at building applications from scratch. Because most users leave software on its default setting, the change could still move substantial usage and spending away from Anthropic even though Claude remains in the product.
Replit previously relied heavily on Claude to power its browser-based development platform, and Masad said the company had recently been among Anthropic’s five largest customers. Anthropic models were still being promoted in Replit material only weeks before the announcement, including references to Claude Fable and Opus in production workflows. The switch illustrates how AI application companies can use a multi-model architecture to negotiate on price and performance without removing alternatives entirely. It also means the practical impact should be measured through future usage, not by treating a preferred-provider decision as an exclusive technical cutoff.