Replit, the AI-driven coding platform favored by developers and startup founders alike, has announced it is discontinuing its partnership with Anthropic. This announcement was made by the company’s CEO and co-founder, Amjad Masad, during his presentation at TechCrunch Disrupt on Thursday.
Masad cited several factors for the decision, including financial considerations. Replit had previously formed a partnership with OpenAI in August, but it was unclear until now that the collaboration with Anthropic had ended.
“I believe ChatGPT has shown significant improvements,” Masad remarked, expressing that the company’s internal benchmarks indicate a preference for using OpenAI’s solutions. “We consistently find that Sol and Astra outperform alternatives in constructing applications from scratch. This decision was not only about cost, but also about performance and the strategic advantages of partnership.”
Replit operates a cloud-based platform allowing users to create and deploy applications directly from their web browsers. Through its previous collaboration with Anthropic, the company utilized the AI lab’s Claude models. This partnership reportedly contributed to a substantial revenue increase for Replit.
However, contradicting Masad’s statements, some insiders have claimed that the partnership with Anthropic remains active. An anonymous source suggested that the collaboration continues.
According to various sources, including the company’s documentation and recent communications, Replit was still utilizing Anthropic products until just weeks ago. A blog post from late September referenced Claude Fable 5 and Claude Fable 5.1 operating in the production environment of Replit Agent. Around this time, a member of the product team also mentioned the use of Opus 5.5, Anthropic‘s leading model, in a post on X.
The TechCrunch team has reached out for further clarification from Replit and will update the story if a response is received.
On stage at TechCrunch Disrupt, Masad also highlighted the company’s growth trajectory. He revealed that enterprise customers now constitute a larger proportion of revenue compared to previous years.
“We aren’t disclosing ARR numbers just yet, but enterprise revenue was around 2% last year. This year, we expect it to rise to approximately 15%, with over 50% of our revenue coming from B2B contracts,” Masad explained.
Masad emphasized the importance of robust AI safety protocols, likening the needs of AI agents to environmental factors that humans require for productivity. He noted, “Agents need a conducive habitat, which includes safety measures, error recovery systems, and efficient network structures for optimal performance.”
This article has been updated to reflect additional insights concerning Replit and its use of Anthropic‘s models.
Editor’s Take
The end of Replit’s partnership with Anthropic raises important questions about the competitive landscape of AI tools used for coding. As Replit pivots towards OpenAI solutions, developers may benefit from potentially superior performance and features. However, the ongoing lack of clarity about Anthropic’s role poses uncertainty for users relying on its models.
Source: techcrunch.com
