Replit, the AI coding platform that has become a favorite of startup founders and developers, is no longer working with Anthropic, the company’s CEO and co-founder Amjad Masad said Thursday while onstage at TechCrunch Disrupt.

Replit Shifts AI Allegiance, Ditching Anthropic for OpenAI Amid Performance & Cost Concerns

In a significant announcement that sent ripples through the AI development community, Replit CEO and co-founder Amjad Masad declared on Thursday at TechCrunch Disrupt that his company is discontinuing its partnership with Anthropic. Replit, a popular cloud-based software development platform, had previously relied on Anthropic’s Claude models to power its AI-driven coding features. This pivot, revealed publicly for the first time, signals a strategic re-evaluation of its core AI infrastructure, moving firmly into the embrace of OpenAI, with whom Replit announced a partnership back in August.

The AI Partnership Pivot: Performance, Cost, and Strategic Alignment

Masad detailed the multifaceted reasons behind the decision to part ways with Anthropic, a move that came as a surprise to many, given the previous success stories attributed to their collaboration. Foremost among the justifications was a direct comparison of AI model efficacy. “I think ChatGPT is better now,” Masad stated candidly, underscoring the platform’s rigorous internal benchmarking process. He further elaborated that Replit’s internal assessments, alongside its publicly available benchmarks, consistently showed OpenAI’s Sol and Astra models outperforming Claude in the creation of “vibe-coded applications from scratch.” This performance differential is critical for Replit, whose platform empowers developers to rapidly build and deploy applications directly from their browsers, making the speed and quality of AI-assisted coding paramount to user experience and productivity.

Beyond raw performance, Masad also cited cost as a pivotal factor in the strategic realignment. For a platform serving millions of developers and increasingly attracting enterprise clients, optimizing operational expenses related to AI model inference is crucial for scalability and profitability. The third pillar of Replit’s decision was an elusive but important concept: “partnership.” This suggests that the relationship with OpenAI extends beyond mere API access, encompassing deeper collaboration, better support, and perhaps a more aligned vision for the future of AI-powered development tools. Masad concluded, “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.” This candid assessment underscores the fierce competition among AI foundational model providers and the strategic choices companies like Replit must make to stay ahead in a rapidly evolving landscape.

Unpacking the Contradiction: Is the Break Immediate?

Despite Masad’s unequivocal statement, the announcement was met with immediate pushback from sources familiar with the Replit-Anthropic relationship. Insiders, who requested anonymity, indicated that the partnership was, in fact, still active. This contradiction introduces an element of ambiguity regarding the exact timeline and nature of Replit’s disengagement from Anthropic. Adding to the confusion, there is tangible public evidence suggesting recent reliance on Anthropic’s models within Replit’s ecosystem. A blog post published as recently as September cited Anthropic’s Claude Fable 5 and Claude Fable 5.1 as integral models operating within Replit Agent’s production environment. Such recent usage directly challenges the notion of an immediate or complete cessation of the partnership.

The discrepancy raises several questions: Is Replit undergoing a phased transition? Was Masad’s statement an announcement of intent rather than a reflection of the current operational reality? Or could there be a lag between strategic decisions and their full technical implementation? The situation highlights the complexities of transitioning core infrastructure, especially when deeply embedded AI models are involved. It also points to potential communication gaps, either between the companies or internally within Replit itself. TechCrunch has reached out to Replit for further clarification on this matter, and the industry awaits an update that could shed more light on the true status of this high-profile AI partnership shift.

Replit’s Expanding Horizons: Enterprise Growth Fuels Ambition

Beyond the headline-grabbing AI partnership news, Masad also shared encouraging insights into Replit’s financial performance and strategic growth trajectory. While refraining from disclosing specific Annual Recurring Revenue (ARR) numbers, the CEO highlighted a dramatic surge in enterprise revenue. Last year, enterprise contracts accounted for a mere 2% of Replit’s total revenue. This year, that figure has skyrocketed to approximately 15%, demonstrating a rapid maturation of Replit’s business model beyond individual developers and startups. Furthermore, Masad noted that more than 50% of the company’s revenue is now derived from B2B (business-to-business) contracts, signaling a robust and expanding footprint within the corporate sector.

This significant shift towards enterprise and B2B clients underscores Replit’s ambition to become an indispensable tool not just for hobbyists and solo developers, but for larger organizations seeking to streamline their software development processes. The ability to cater to enterprise needs often requires more robust security, compliance features, dedicated support, and scalable infrastructure – all areas where Replit is clearly making strides. This growth trajectory is critical for Replit’s long-term sustainability and potential valuation, positioning it as a serious contender in the competitive cloud development and AI tooling market. The move to a potentially “better partner” like OpenAI could also be seen as a strategic play to further enhance features critical for enterprise adoption, such as advanced code generation, debugging, and security analysis capabilities.

Pioneering AI Safety: Crafting a “Habitat” for Agents

In a forward-looking discussion, Masad also addressed the crucial topic of AI safety, emphasizing its foundational importance for the responsible deployment of AI agents. He drew an insightful analogy: “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.” This “habitat,” as Masad described it, is a comprehensive ecosystem designed to enable AI agents to function effectively while mitigating risks.

He outlined the essential components of such a habitat: good orchestration for managing complex AI tasks, robust error recovery mechanisms to handle unexpected issues, stringent safety and security protocols, effective sandboxing to isolate potentially harmful operations, reliable networking for communication, and strong identity and authorization controls. These elements are not merely technical specifications; they represent a holistic approach to ensuring AI systems operate predictably, securely, and ethically within a development environment where users might be experimenting with various applications, some of which could interact with sensitive data or systems. Replit’s focus on building this secure “habitat” reflects a broader industry recognition that as AI agents become more autonomous and integrated into workflows, the infrastructure supporting them must be designed with safety and control as paramount considerations.

The Bottom Line

Replit stands at a pivotal juncture, making bold strategic decisions that underscore its ambitions in the rapidly evolving AI and cloud development landscape. The announced shift from Anthropic to OpenAI, driven by perceived superior performance, cost efficiencies, and partnership alignment, highlights the intense competition among foundational AI model providers and the critical choices faced by platforms integrating these technologies. While the immediate nature of this split remains contentious, with conflicting reports from insiders, Replit’s trajectory is clearly pointing towards greater enterprise adoption and a maturing business model. Coupled with a strong emphasis on building a secure and controlled “habitat” for AI agents, Replit is positioning itself not just as an innovator in AI-powered coding, but also as a responsible steward of emerging technologies. The coming months will reveal how seamlessly this AI transition unfolds and how effectively Replit leverages its new strategic partnerships to further solidify its market position and navigate the complex demands of both individual developers and growing enterprise clientele.