Replit Makes OpenAI Its Default Model Provider

Replit is changing the model mix behind its browser-based coding platform. CEO and co-founder Amjad Masad said OpenAI has replaced Anthropic as Replit’s default model provider. The shift does not remove Claude from the service: customers can still select Anthropic models and other supported options.

Key points

• OpenAI is now the default. Replit announced a partnership with OpenAI in August. Masad said the decision was based on internal and public benchmarking, with OpenAI models performing better for building applications from scratch in the company’s testing. Cost and the quality of the partnership were also factors.
• Claude remains available. Replit continues to present itself as model agnostic. OpenAI is the preferred partner for powering Replit Agent, but users can still choose Anthropic models. Because most users stay with the default, the change could nevertheless reduce Replit’s spending with Anthropic.
• The status of the Anthropic relationship is unclear. Masad said onstage that Replit was no longer working with Anthropic. An unnamed source familiar with the partnership disputed that characterization, and recent Replit pages, blog posts, and employee social posts continued to reference Anthropic models. The clearest interpretation is a change in default provider, not necessarily a complete termination.
• Enterprise is becoming more important. Masad said enterprise accounted for about 2% of revenue last year and roughly 15% this year, while more than half of revenue now comes from B2B activity. He expects enterprise and government customers to represent more than half of revenue in 2027.

Why it matters

For an AI coding platform, the default model is both a technical and commercial decision. It affects inference costs, user experience, product behavior, and negotiating leverage with model vendors. Replit’s move from relying heavily on Claude to prioritizing OpenAI shows how model competition is increasingly being fought through distribution: the provider behind the default setting is likely to receive the largest share of routine traffic and product feedback.

Keeping Claude available also reflects the limits of a one-model strategy. Developers may prefer different systems for code generation, task planning, reliability, or price. Supporting multiple providers can reduce dependence on any single vendor, but it requires more routing, evaluation, and cost controls.

Masad additionally argued that agents need a suitable “habitat” to work safely. That includes orchestration, error recovery, sandboxing, networking, identity, and authorization. As Replit expands toward enterprise and government customers, those controls may become as important as raw model capability. The impact of the provider switch will ultimately depend on how well Replit validates model performance in real development workflows.

Source: TechCrunch AI