Replit has made OpenAI its preferred partner for powering Replit Agent, but Anthropic's Claude models remain available. CEO Amjad Masad clarified that distinction to TechCrunch after saying onstage at Disrupt that the coding platform was no longer working with Anthropic.

Masad attributed the preference to performance, economics and the partnership itself. His clarification described Replit as model agnostic, with users still able to choose Claude or other supported models. The announcement therefore does not establish a complete removal of Anthropic from the service.

What users can actually select

Replit's current model-selector documentation lists both OpenAI and Anthropic models. Core and Pro users can manually choose an available primary model in Power or Max mode. Free Mode instead uses intelligent routing and does not offer manual model selection.

Availability is qualified: Replit says the selector is rolling out, and options can vary with organization settings and authorization. The model list inside an individual account is the authoritative guide for that user.

The company's engineering blog also describes Claude Fable models and GPT-6 Astra operating in Replit Agent production, providing additional context against interpreting the change as an outright Claude ban.

Preference is not exclusivity

For developers, the practical distinction is between a platform's preferred provider and the models it exposes. Teams that require a particular model should check account-level controls rather than assume the default guarantees it.

Masad's performance assessment remains a company claim, not an independent finding that OpenAI is better for every coding workload. The verified change is Replit's stated provider preference, alongside continued documented Claude access.