Replit Switches Default AI Model to OpenAI from Anthropic

Image: TechCrunch · Source

Replit now defaults to OpenAI's ChatGPT for its AI coding platform, citing better performance, cost, and partnership strength, while still allowing access to Anthropic models.

Replit, the cloud-based coding platform favored by developers and startups, has made OpenAI’s ChatGPT its default AI model, replacing Anthropic’s Claude as the primary engine powering its services. This strategic shift was confirmed by Replit CEO and co-founder Amjad Masad at TechCrunch Disrupt on October 16, 2026.

The decision factors in multiple considerations. Masad explained that OpenAI’s models demonstrated superior performance on Replit’s internal benchmarks, especially in building code-centric applications from scratch. Additionally, OpenAI’s offerings presented more favorable economics compared to Anthropic, alongside stronger existing partnership ties. Collective evaluation of performance, cost, and collaboration led Replit to designate OpenAI as its preferred AI provider for powering Replit Agent, the platform’s AI assistant.

Despite the switch in default, Replit remains model agnostic. Customers can still select Anthropic’s Claude and other supported AI models on the platform. However, since most users tend to use the default option, the change is likely to reduce Replit’s usage volume and spending on Anthropic’s services significantly.

Before partnering with OpenAI, Replit heavily relied on Anthropic’s Claude models and had grown to become one of Anthropic’s top five customers. The partnership had previously contributed notably to Replit’s revenue scale.

Masad provided insights into Replit’s evolving business mix, highlighting a marked increase in enterprise revenue. Last year, enterprise customers contributed roughly 2% of revenue; this year, that figure has risen to 15%, with business-to-business sales now comprising over half of total income. Masad anticipates this trend to continue with enterprise and government clients expected to deliver more than 50% of revenue by 2027.

On AI safety, Masad emphasized the importance of a robust operational environment for AI agents, akin to a conducive workplace for humans. Proper orchestration, error recovery, security, sandboxing, and identity management collectively create a safe and effective AI habitat, essential for reliable deployment.

While some insiders have contested the notion that the partnership with Anthropic has ended, and Replit’s public materials indicate ongoing support for Anthropic models, the designation of OpenAI as the default confirms a significant pivot in the company’s AI strategy.

This update aligns with Replit’s efforts to optimize performance and cost efficiency while maintaining flexibility and choice for users within its AI-augmented coding platform.