During a session at TechCrunch Disrupt, Replit chief executive Amjad Masad disclosed that the company will now present OpenAI's ChatGPT as the default model for its cloud-based development environment. The announcement follows a partnership with OpenAI that was first revealed in August, and marks a shift from the earlier reliance on Anthropic's Claude series.
Masad explained that the change was driven by a combination of cost considerations, partnership dynamics, and performance outcomes. He referenced an internal benchmark used to evaluate metrics such as speed and accuracy, noting that the public benchmark consistently shows Replit's Sol and Astra models outperforming alternatives on new code generation tasks.
Although OpenAI will be the pre-selected option, Anthropic's models will still be accessible to users who manually choose them, a detail Masad clarified after the stage presentation. Because most developers tend to accept the default setting, Replit anticipates a notable reduction in its spending on Anthropic services.
Masad emphasized that Replit continues to maintain a model-agnostic stance, allowing developers to opt for Claude or any other supported model. He told TechCrunch that the decision to prioritize OpenAI for the Replit Agent stemmed from the current balance of performance, economics, and the strength of the partnership.
Insiders familiar with the arrangement suggested that the Anthropic collaboration has not been fully terminated, pointing to recent blog entries and social-media posts that still highlight Claude Fable 5, Claude Fable 5.1, and the Opus 5.5 model as active components of Replit Agent's production stack.
Masad also shared revenue trends, noting that enterprise contracts grew from roughly two percent of total income last year to about fifteen percent this year, while business-to-business sales now exceed fifty percent. He projected that more than half of Replit's 2027 revenue will derive from enterprise and government customers, underscoring the importance of robust AI agent infrastructure.