Replit, the popular cloud-based software development platform that has become a staple for startup founders and independent developers, is officially parting ways with AI powerhouse Anthropic. The announcement, delivered by Replit CEO and co-founder Amjad Masad during an onstage appearance at TechCrunch Disrupt this Thursday, marks a significant pivot in the platform’s underlying infrastructure strategy. This decision comes just months after Replit publicly heralded a new partnership with OpenAI in August, signaling a broader consolidation of the company’s AI roadmap.
The transition, according to Masad, is rooted in a multifaceted evaluation of current market offerings. While the collaboration with Anthropic—which saw Replit integrating the lab’s sophisticated Claude models to power its development environment—was widely credited with driving a tenfold increase in the company’s revenue, the landscape of generative AI has shifted rapidly. For Replit, the decision to move away from Anthropic was not a singular reaction to one factor, but rather a holistic assessment encompassing cost efficiency, technical performance, and strategic alignment.
"I think ChatGPT is better now," Masad stated during the Disrupt session, providing a blunt rationale for the shift. He explained that Replit maintains a rigorous internal benchmarking system designed to evaluate the real-world performance of various large language models. Beyond their internal testing, the company also publishes public-facing benchmarks to provide transparency into how their tools, specifically the "Sol" and "Astra" models, perform under stress.
"We publish a public benchmark as well, and consistently, Sol and Astra perform better on building vibe-coded applications from scratch," Masad continued. "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."
A Disputed Narrative
Despite Masad’s definitive comments, the situation appears more fluid than a clean break might suggest. Following his announcement, reports surfaced from insiders familiar with the relationship between the two companies, suggesting that the partnership may not be as definitively concluded as the public remarks imply. One source, speaking on the condition of anonymity, indicated that the collaboration between Replit and Anthropic remains in place, contradicting the notion of an immediate and total severance.
There is tangible evidence that suggests a continued technical dependency on Anthropic’s ecosystem. As recently as September, a blog post published on Replit’s own website explicitly cited the use of Anthropic’s "Claude Fable 5" and "Claude Fable 5.1" models within the production environment of Replit Agent, the company’s flagship autonomous coding assistant. This discrepancy between the CEO’s public stance and the company’s documented technical operations has created a degree of uncertainty. In response to these conflicting signals, TechCrunch has reached out to Replit for formal clarification, though at the time of writing, no further comment had been provided.
The Evolution of the Replit Platform
Replit has carved out a unique niche in the software development space by offering a cloud-native platform that allows users to write, test, and deploy applications directly within their web browsers. By lowering the barrier to entry, it has become an essential tool for "vibe-coding"—a term that captures the current era of rapid, AI-assisted application building.
The previous integration with Anthropic was a cornerstone of this experience. When the partnership was first established, it provided Replit with high-performance, reasoning-heavy models that were instrumental in helping the platform scale its user base and revenue. The transition to OpenAI’s models, therefore, represents more than just a vendor swap; it reflects a change in how Replit intends to deliver value to its users as the complexity of AI-generated code continues to grow.
Scaling and the Enterprise Shift
Beyond the headlines regarding its AI model providers, Masad’s appearance at TechCrunch Disrupt offered a rare glimpse into the health and trajectory of Replit’s business. As the company matures, its revenue composition is undergoing a notable transformation. While Replit has historically been defined by its massive community of individual developers and hobbyists, the company is seeing a significant influx of enterprise interest.
"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 shared with the audience. This figure has expanded rapidly, with the CEO noting that, "This year, I would say like 15% of our revenue is enterprise contracts, and more than 50% of our revenue is B2B."
This shift toward enterprise contracts indicates that businesses are increasingly relying on Replit’s browser-based environment to manage their development workflows, security, and deployment pipelines. This move into the enterprise sector necessitates a higher degree of stability and predictability, which likely influenced the decision to re-evaluate the platform’s underlying AI model partnerships.
The Future of AI Agents and Infrastructure
The conversation at Disrupt also touched upon the broader philosophy of AI deployment, specifically regarding the "habitat" required for AI agents to function effectively. Masad emphasized that simply having access to a high-performing model is insufficient for professional-grade software development. He likened the needs of an AI agent to those of a human worker, arguing that for an agent to perform consistently, it must exist within a robust, well-defined ecosystem.
"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. He underscored that this habitat must provide more than just raw compute; it requires sophisticated orchestration, error recovery, rigorous safety and security protocols, sandboxing, networking, and robust identity and authorization management.
According to Masad, these infrastructure layers are the true differentiators in the current AI market. "Once you figure all of that out, then you can deploy AI safely," he said. This focus on "AI safety" and "habitat" design underscores why Replit might be prioritizing its own internal models like Sol and Astra, as well as its strategic relationship with OpenAI. By controlling more of the orchestration layer, Replit aims to provide a more secure and predictable environment for its users, particularly those in the enterprise sector who require strict adherence to security and compliance standards.
As Replit continues to navigate the volatile landscape of generative AI, its ability to balance the rapid, iterative pace of consumer-facing features with the stability required for enterprise clients will remain critical. Whether the relationship with Anthropic is truly over or merely entering a new, less integrated phase remains to be seen. For now, the company’s focus remains clear: building a "habitat" that can sustain the next generation of autonomous coding, regardless of which underlying model is currently providing the intelligence. As the company prepares for its next stage of growth, these strategic choices regarding its model partners and its focus on enterprise-grade infrastructure will serve as the foundation for its future in an increasingly crowded and competitive AI development market.
