Replit, the AI coding platform favored by many startup founders and developers, has ceased its collaboration with Anthropic, according to remarks made by CEO and co-founder Amjad Masad during his appearance at TechCrunch Disrupt on Thursday.
Masad highlighted that the decision stemmed from multiple factors, with cost being a significant consideration. While Replit had previously announced a partnership with OpenAI in August, there had been no prior indication of the termination of its relationship with Anthropic.
“I think ChatGPT is better now,” Masad commented, noting that the company employs internal benchmarks for performance assessment. “We also publish a public benchmark, and consistently, our tools perform better when it comes to developing vibe-coded applications from scratch. This decision was driven by cost, partnership dynamics, and overall performance. Given these factors, we determined that OpenAI is the better partner for us at this time.”
Interestingly, while Replit is shifting to prioritize OpenAI’s services, it appears that Anthropic’s tools are not entirely phased out. Following the Disrupt conference, Masad clarified that users can still access Anthropic models on the platform, but OpenAI’s ChatGPT will now be the default choice. This strategic move could significantly decrease Replit’s expenses related to Anthropic, as users typically prefer default models over manually selecting alternatives.
Prior to partnering with OpenAI, Replit utilized Anthropic’s Claude models, a collaboration that reportedly led to a tenfold revenue increase for the company. Masad stated that until recently, Replit ranked among Anthropic’s top five customers.
“Replit remains model agnostic,” Masad emphasized. “Our users have the option to choose Anthropic’s Claude or other supported models, but OpenAI is now our preferred partner for powering Replit Agent due to performance, cost-effectiveness, and the strength of our partnership.”
However, some individuals familiar with the previous collaboration indicated that the partnership between Replit and Anthropic remains intact. Evidence available on Replit’s website and social media as recently as a few weeks ago shows active promotion of Anthropic’s products. A blog post from late September, for instance, mentioned models like Claude Fable 5 and Claude Fable 5.1 operating within Replit Agent’s ecosystem.
Furthermore, Masad provided insights into Replit’s growth trajectory during the Disrupt event, indicating that enterprise revenue is becoming increasingly significant.
“We’re withholding our annual recurring revenue figures for now, but to give you a sense of scale, enterprise contributed about 2% of our revenue last year,” he noted. “This year, it’s approximately 15%, and more than half of our revenue now comes from B2B contracts.”
Masad anticipates that by 2027, more than half of Replit’s revenue will be derived from enterprise and government clients.
He also discussed the importance of robust AI safety protocols, stating, “Just like humans require a conducive workspace—comprising good food, air conditioning, and a reliable computer—AI agents too need an optimal environment to function effectively. This involves orchestrated safety measures, error recovery processes, and excellent networking and authorization systems. Once these elements are in place, AI can be deployed securely.”
This article was originally published on October 16 at 2:22 pm PT and has been updated to provide additional context regarding Replit’s partnerships.
Editor’s Take
This development highlights a pivotal shift in the AI landscape, revealing how partnerships can affect the tools available to developers. By prioritizing OpenAI, Replit may enhance its platform’s capabilities and cost efficiency while continuing to offer Anthropic’s models, thus maintaining flexibility for users. Observing how this strategy evolves will be crucial for users, developers, and the broader AI ecosystem.
Source: techcrunch.com
