Replit, the popular AI-powered coding platform that has steadily earned a devoted following among startup founders and software developers alike, is no longer working with AI lab Anthropic. The surprising revelation was made by Amjad Masad, the CEO and co-founder of Replit, while speaking onstage at TechCrunch Disrupt.

Addressing an audience of industry leaders, technologists, and entrepreneurs, Masad explained that the decision to sever ties with Anthropic was driven by a combination of factors, chief among them being cost considerations, shifting partnership dynamics, and performance evaluations. The announcement sheds new light on Replit’s strategic evolution, particularly coming several months after the company publicly announced a major partnership with OpenAI in August. Until Masad’s remarks at Disrupt, however, there had been no public indication that Replit’s underlying technical infrastructure had fully parted ways with Anthropic.

Elaborating on the performance metrics that guided the company’s decision-making process, Masad pointed directly to OpenAI’s current technological capabilities. “I think ChatGPT is better now,” Masad told the audience, noting that Replit continuously relies on a rigorous internal benchmarking system to assess key performance metrics across various large language models. “We publish a public benchmark as well, and consistently, Sol and Astra perform better on building vibe-coded applications from scratch. It was a decision based on both cost, partnership, and performance. And for all these reasons, we just felt like they were a better partner for now.”

For those following the trajectory of cloud-based development tools, Replit has established itself as an essential software development platform where users can write, develop, test, and deploy full-fledged applications directly from inside their web browsers. In earlier phases of its modern AI integration, Replit heavily leveraged Anthropic’s advanced Claude models to power core platform features, including its automated coding agents. Industry observers and financial reports have frequently noted that this high-profile partnership with Anthropic served as a major catalyst for Replit, significantly accelerating its user acquisition and helping the enterprise grow its revenue exponentially over the past year.

Despite Masad’s clear and definitive statements onstage at TechCrunch Disrupt, the situation appears to carry some internal complexity. Shortly after the remarks were made, comments from the CEO were met with some pushback from industry insiders who are closely familiar with the operational history and ongoing commercial agreements between Anthropic and Replit. One individual familiar with the partnership, who spoke on the condition of anonymity due to the sensitive nature of corporate alliances, maintained that the partnership between the two companies was still in place.

Furthermore, tangible evidence indicates that Replit was actively utilizing Anthropic products in its production environments as recently as September. Documentation from a company blog post published in September explicitly cites Anthropic’s Claude Fable 5 and Claude Fable 5.1 models as operating directly within the production environment of Replit Agent, the platform’s autonomous software engineering assistant. In light of these conflicting accounts, TechCrunch reached out to Replit for official clarification regarding the timeline and current legal status of the vendor relationship, promising to update its reporting should a formal response be provided.

Beyond the shifting landscape of its underlying AI model providers, Masad used his time onstage at TechCrunch Disrupt to pull back the curtain on Replit’s broader business growth and financial evolution. The platform’s leadership has traditionally kept detailed financial disclosures private, but Masad offered a revealing look at how the company’s revenue streams have matured, highlighting a dramatic pivot toward enterprise monetization.

“We’re not sharing ARR numbers just yet, but in terms of growth, like, I would say, enterprise was like 2% of revenue last year,” Masad noted, illustrating the modest footprint that large corporate contracts held in the company’s early days. That reality has transformed remarkably over the course of a single year. “This year, I would say like 15% of our revenue is enterprise contracts, and more than 50% of our revenue is B2B.”

The conversation at Disrupt also turned toward the broader operational realities of deploying autonomous artificial intelligence in real-world production environments. Addressing the growing industry discourse around AI safety, governance, and reliability, Masad drew a parallel between human labor environments and the infrastructural requirements needed for software-writing agents to thrive.

“In the same way that us humans need a certain environment to work — you need a nice office, maybe nice good food in the office, air-conditioning, computer — agents need a habitat for them to exist and do good work,” Masad explained to the Disrupt audience. He emphasized that building this digital habitat requires deliberate architectural foresight rather than merely plugging an LLM into a blank terminal. “So that habitat has good orchestration, good error recovery, good safety and security and sandboxing, good networking, good identity and authorization. And once you figure all of that out, then you can deploy AI safely.”