Replit CEO Shocks Tech World: Partnership with Anthropic Ceases

Replit, a prominent AI coding platform favored by startup founders and developers, has significantly shifted its primary artificial intelligence partnership from Anthropic to OpenAI. This strategic decision was publicly announced by Replit CEO and co-founder Amjad Masad during TechCrunch Disrupt, citing a confluence of factors including cost, partnership dynamics, and performance metrics. While Replit had previously established a partnership with OpenAI in August, the transition away from Anthropic had not been explicitly communicated until this announcement.

Masad elaborated on the rationale behind the change, stating, "I think ChatGPT is better now." He emphasized that Replit employs both internal and public benchmarks to rigorously assess performance, noting that OpenAI's Sol and Astra models consistently outperform others in building AI-coded applications from inception. Consequently, the decision was a comprehensive one, factoring in economic efficiency, the strength of the partnership, and the superior performance observed from OpenAI's offerings. This led Replit to conclude that OpenAI was a more suitable partner for the immediate future.

Despite this pronouncement, the relationship with Anthropic has not been entirely severed. Masad clarified post-Disrupt that while OpenAI’s ChatGPT is now the default model on Replit’s cloud-based software development platform, customers retain the ability to access Anthropic models. This change, however, is expected to substantially reduce Replit's expenditures with Anthropic, given that most users typically opt for the default settings rather than manually selecting alternative AI models.

Historically, Replit's platform relied heavily on Anthropic's Claude models prior to securing its agreement with OpenAI. This previous collaboration was instrumental in Replit's significant revenue growth, with Masad acknowledging that Replit had, until recently, been among Anthropic's top five customers. Despite the shift in primary partnership, Masad underscored Replit's commitment to being "model agnostic," telling TechCrunch that "Our users can choose Anthropic’s Claude or other supported models…OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.”

However, Masad's initial comments regarding the partnership were met with skepticism from insiders familiar with the Anthropic-Replit relationship. One anonymous source indicated that the partnership was, in fact, still active. Evidence supporting this view includes Replit's recent promotion of Anthropic products across its website, blog posts, and social media channels. For example, a late September blog post specifically mentioned Anthropic’s Claude Fable 5 and Claude Fable 5.1 as active models within Replit Agent's production environment. Around the same period, a Replit product employee publicly promoted Anthropic’s flagship Opus 5.5 model on X, highlighting its capabilities.

Beyond the AI partnership discussions, Masad also provided insights into Replit's broader business growth. He noted a significant increase in enterprise revenue, stating that while enterprise contracts constituted approximately 2% of revenue last year, they now account for about 15%, with over 50% of total revenue derived from B2B operations. Looking ahead, Masad anticipates that by 2027, more than half of Replit's revenue will originate from enterprise and government clientele.

Masad further addressed the critical importance of robust AI safety measures. He drew an analogy, explaining that just as humans require a conducive environment to perform effectively, AI agents need a "habitat" for optimal and safe operation. This habitat, according to Masad, must encompass strong orchestration, efficient error recovery, comprehensive safety and security protocols, effective sandboxing, reliable networking, and robust identity and authorization systems. Once these foundational elements are in place, AI can be deployed securely and responsibly.