Replit CEO Says Platform No Longer Works With Anthropic

Replit CEO and co-founder Amjad Masad said Thursday at TechCrunch Disrupt that the AI coding platform is no longer working with Anthropic, signaling a shift in the company’s model partnerships after a widely publicized collaboration that had helped scale the business. Masad said the decision was based on cost, partnership terms and performance, adding that Replit’s internal and public benchmarks now favored OpenAI models. The remarks follow Replit’s August partnership announcement with OpenAI and its prior use of Anthropic’s Claude models to power its browser-based development and deployment platform. Masad’s account drew pushback from people familiar with the arrangement, some of whom said the Anthropic partnership remained in place. The uncertainty is compounded by references in Replit materials as recently as September to Anthropic’s Claude Fable 5 and Claude Fable 5.1 models running in Replit Agent’s production environment. Replit did not immediately provide clarification on whether the relationship had formally ended or was being restructured. The issue sits at the center of Replit’s business model, which sells a cloud-based software development environment where users can build, test and deploy applications from a browser. AI models underpin core features such as code generation, application scaffolding and agent-driven workflows. For Replit, the choice of model provider affects performance, cost structure, platform capability and its positioning against competitors in AI-assisted coding and developer tooling. Masad also used the session to describe how Replit’s commercial base is changing. He said enterprise now accounts for about 15% of revenue, up from roughly 2% last year, and that more than 50% of revenue comes from business-to-business customers. He did not disclose annual recurring revenue, but the figures suggest a broadening move toward enterprise adoption as Replit expands beyond individual developers and startup users. Masad also discussed the infrastructure required to deploy AI agents safely. He described agents as needing a controlled operating environment with good orchestration, error recovery, safety, security, sandboxing, networking, identity management and authorization controls. The framing reflects a broader industry trend: coding assistants and autonomous agents are increasingly expected to operate inside governed software platforms rather than as open-ended chat tools, with stronger permissions management, observability and containment around their actions. For Replit, the reported distancing from Anthropic could have practical consequences for platform reliability, model availability, enterprise trust and pricing. Anthropic’s Claude models have been central to Replit’s public narrative and product performance, and any change in that relationship may require retooling around OpenAI or other model providers. It also underscores the volatility of AI model partnerships, where performance benchmarks, cost negotiations and strategic positioning can change quickly and become visible through public remarks. The exchange highlights both opportunity and risk for AI-native developer platforms. Strong model partnerships can accelerate growth and differentiate products, but dependence on a single provider can create operational and commercial exposure. Replit’s enterprise momentum, combined with its reliance on foundation models, places the company at the intersection of developer productivity, model competition and enterprise security expectations.