Replit has designated OpenAI as its primary default AI model provider instead of Anthropic to boost performance and cut platform costs, its CEO said.
Speaking at TechCrunch Disrupt, Replit chief executive officer and co-founder Amjad Masad revealed that the cloud-based coding platform is no longer relying on Anthropic as its primary model partner. Instead, ChatGPT creator OpenAI has taken over as the preferred provider powering Replit Agent, following an initial deal struck between Replit and OpenAI earlier in August.
According to reporting by TechCrunch, Masad explained that the strategic decision was driven by a combination of performance evaluation, financial economics, and general partnership alignment. He noted that ChatGPT currently delivers superior results for building applications from scratch, pointing to internal metrics and public benchmarks where OpenAI models Sol and Astra outperformed competing alternatives in vibe-coded app creation.
Clarifying the arrangement following the event, Masad emphasized that Replit has not completely severed ties with Anthropic or removed its models from the ecosystem. While Replit maintains a model-agnostic stance allowing customers to manually choose Anthropic's Claude or other supported models, OpenAI's offerings now serve as the platform's default setting. Because the vast majority of software developers typically utilize standard platform defaults, this transition is anticipated to sharply decrease Replit's overall operational expenditure with Anthropic.
The announcement comes after Replit previously spent months relying heavily on Anthropic's Claude models to drive its core development tools. That initial collaboration fueled significant revenue expansion for Replit, which had recently ranked among Anthropic's top five largest client accounts.
However, TechCrunch reported that people familiar with the matter rejected the characterization that the partnership between Anthropic and Replit had dissolved, asserting that the arrangement remains active. Recent activity across Replit's promotional channels supports this continued technical availability. In late September, official Replit blog posts mentioned running Anthropic's Claude Fable 5 and Claude Fable 5.1 within Replit Agent production environments, while staff publicly highlighted the capabilities of Anthropic's flagship Opus 5.5 model.
Alongside updates on model partnerships, Masad offered insights into Replit's financial trajectory and expanding commercial footprint. Although the company declined to reveal exact annual recurring revenue figures, enterprise contracts have expanded dramatically. Enterprise accounts accounted for just 2 percent of total revenue last year, but have scaled to roughly 15 percent this year, with overall business-to-business clients generating more than 50 percent of total revenue. Looking ahead, Masad stated he anticipates enterprise and government organizations will account for over half of Replit's revenue by 2027.
Addressing infrastructure and safety requirements for autonomous coding systems, Masad framed AI agents as needing tailored technical environments to function effectively. He stated that agents require structured habitats equipped with robust orchestration, error recovery, sandboxing, networking, and identity authorization to ensure safe deployment.
Background
Replit operates a popular cloud-based software development platform that allows programmers and startup founders to build, test, and deploy software applications directly inside a web browser. Over the past year, the company introduced autonomous tools like Replit Agent to streamline software creation through artificial intelligence.
To power these automated workflows, Replit initially built its core offerings on top of Anthropic's Claude frontier models. That technical integration coincided with rapid financial growth for Replit and turned the company into one of Anthropic's top five accounts. In August, Replit expanded its infrastructure footprint by forming a new strategic partnership with OpenAI, setting the stage for the recent shift in default model selection.
Why It Matters
The operational shift by Replit illustrates the intense competition taking place among leading artificial intelligence foundation model developers to secure platform defaults. Because default system configurations drive the vast majority of end-user consumption, winning default status on high-traffic platforms like Replit directly impacts API query volume and recurring infrastructure revenue.
Furthermore, Replit's emphasis on balancing model performance against underlying compute costs highlights growing economic discipline among AI startups. As platforms scale their enterprise and business-to-business offerings, optimizing margins while maintaining competitive software generation capabilities becomes essential. The change also demonstrates how multi-model architecture allows developer platforms to retain flexibility, leveraging competing model providers based on changing benchmarks and commercial terms without entirely cutting off developer access to alternative ecosystems.
Illustration: AI-generated image, not a photograph of the event.
Source