Amjad Masad, CEO and co-founder of the Replit platform, announced that OpenAI has become the platform’s default model provider instead of Anthropic. Masad said the decision was based on a combination of factors, including performance, cost, and the strength of the partnership, after Replit’s internal tests concluded that OpenAI’s models performed better at building applications from scratch in AI-assisted coding scenarios.
Replit announced a partnership with OpenAI in August, but the announcement at the time did not clearly indicate that its cooperation with Anthropic had ended. Masad later explained that platform users could still choose Claude and other supported models, while ChatGPT had become the default option for powering Replit Agent.
A Default Change That Could Redistribute Spending
Replit allows users to develop and deploy applications from within the browser, meaning that the choice of default provider practically affects the models most customers will use. According to the report, many users tend to stick with the default option rather than change it manually, which could reduce Replit’s spending on Anthropic models even though they remain available on the platform.
Replit previously relied on Claude models, and Masad said that until recently the company was among Anthropic’s five largest customers. He also noted that the partnership had helped Replit’s revenue grow significantly before OpenAI became the preferred partner for powering Replit Agent.
Conflicting Indicators About the Status of the Partnership
The boundaries of the shift do not appear to be entirely clear. The report quoted a person familiar with the relationship, who requested anonymity, as saying that the partnership between Replit and Anthropic was still in place. Replit also continued promoting Anthropic’s products; a blog post published in late September referred to the use of Claude Fable 5 and Claude Fable 5.1 in Replit Agent’s production environment, while one of the product employees promoted the Opus 5.5 model around the same time.
Therefore, the confirmed change is OpenAI’s move into the position of default and preferred provider, not necessarily Anthropic’s removal from the technical infrastructure or commercial relationship.
Why Does This News Matter?
The decision reflects the reliance of AI coding platforms on balancing three factors simultaneously: model quality for specific tasks, operating costs, and partnership terms. It also shows that changing the default setting can alter the distribution of usage and revenue among model providers without preventing users from accessing alternatives.
This coincided with the expansion of the share of businesses and enterprises in Replit’s operations. Masad said enterprise contracts accounted for about 2% of revenue last year and rose to about 15% this year, while business revenue exceeded 50% of the total. More than half of revenue is expected to come from enterprise and government customers in 2027.
Masad also stressed that deploying AI agents safely requires an operating environment that includes task orchestration, error recovery, security isolation, networking, and identity and authorization management. The details of the current commercial agreement between Replit and Anthropic, as well as the extent of the actual shift in usage, have not been fully disclosed.