Replit Ends Anthropic Partnership, Shifts to OpenAI Models

Replit has officially ended its collaboration with AI lab Anthropic, according to the cloud-based coding platform's chief executive Amjad Masad. Speaking at TechCrunch Disrupt, Masad cited performance metrics and cost efficiency as primary drivers for the shift toward OpenAI, with whom Replit began a partnership in August.

Performance and Benchmarks

Masad noted that internal evaluations and public benchmarks now favor OpenAI models like Sol and Astra for building complex applications from scratch. He described the transition as a strategic move to secure a more effective partner for the platform's current trajectory. The Scale Report understands that despite this public declaration, the technical transition remains a point of contention, as anonymous insiders have suggested that the relationship with Anthropic may still be active in some capacity.

Recent Technical Usage

Evidence of Anthropic integration persists on the Replit website and social media channels. As recently as late September, official blog posts identified Claude Fable 5 and 5.1 as primary engines within the Replit Agent environment. Additionally, Replit staff members were promoting the capabilities of Opus 5.5 on social platforms throughout the same period, casting uncertainty on the timeline of the alleged termination.

Enterprise Growth

Beyond the model swap, Masad provided an update on the company's financial evolution, highlighting a pivot toward larger institutional clients. Enterprise contracts have expanded from 2 percent of revenue last year to 15 percent today, with business-to-business services now accounting for more than half of the company's total income. Masad emphasized that providing a robust environment for AI agents, including secure sandboxing and identity authorization, remains a core focus for the company as it scales.