TL;DR
- Default Provider Switch: Replit has designated OpenAI as its default model provider, replacing Anthropic to power its developer tools and the autonomous Replit Agent.
- Access Retained: Despite initial onstage remarks by CEO Amjad Masad stating Replit was "no longer working with Anthropic," Masad later clarified that Replit remains model agnostic and users can still select Anthropic's Claude.
- Selection Criteria: The transition was driven by cost considerations, partnership factors, and benchmark results, with OpenAI models including Sol and Astra outperforming competitors in generating vibe-coded applications from scratch.
- Partnership Dispute: Insiders familiar with the relationship rejected claims of a full split, stating the commercial partnership remains active, consistent with recent promotions of Claude Fable and Opus 5.5 on Replit's channels.
- Enterprise Growth: Replit reported significant enterprise momentum, with enterprise contracts rising from 2% to 15% of total revenue year-over-year and B2B contracts now exceeding 50% of revenue.
Replit, the cloud-based development environment widely adopted by startup founders and software engineers, has transitioned to OpenAI as its default model provider. The strategic shift follows statements made by Replit co-founder and Chief Executive Officer Amjad Masad during the TechCrunch Disrupt conference, accompanied by subsequent clarifications regarding the company's relationship with former primary provider Anthropic.
The platform provides a browser-based environment where developers build, test, and deploy applications without local configuration. Before formalizing a partnership with OpenAI in August, Replit relied primarily on Anthropic's Claude models to drive its coding infrastructure. According to Masad, Replit had until recently been a top-five customer of Anthropic, a commercial arrangement that contributed significantly to Replit's revenue growth.
Onstage Announcement and Subsequent Clarifications
Speaking onstage at TechCrunch Disrupt, Masad initially told attendees that Replit was no longer working with Anthropic, citing multiple operational reasons, including computational cost. While Replit had announced a partnership with OpenAI in August, the company had not previously indicated an end to its collaboration with Anthropic.
Following the conference, however, Masad clarified the nature of the platform's model routing. Rather than completely removing Anthropic, Replit has established OpenAI's ChatGPT as its default model provider. Masad emphasized that the platform maintains support for external architectures:
“Replit remains model agnostic,” Masad told TechCrunch over the weekend. “Our users can choose Anthropic’s Claude or other supported models…OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.”
In practice, the designation of a default model provider carries substantial economic implications. Because the majority of developers utilize default settings rather than manually configuring custom models, routing primary traffic to OpenAI is expected to reduce Replit's compute expenditure with Anthropic.
Contrasting Accounts of the Anthropic Partnership
The characterization that Replit had severed working ties with Anthropic was met with disagreement from individuals familiar with the relationship. Insiders rejected Masad's onstage comments, with one person speaking on condition of anonymity to state that the partnership between Anthropic and Replit remains in place.
Public activity across Replit's platforms leading up to the conference also indicated ongoing integration. In late September, an official Replit blog post cited Anthropic's Claude Fable 5 and Claude Fable 5.1 as active models within Replit Agent's production environment. Around the same time, a Replit product employee posted publicly on X highlighting the capabilities of Anthropic's flagship Opus 5.5 model.
Performance Benchmarks and 'Vibe-Coded' Applications
Discussing the rationale behind designating OpenAI as the preferred partner, Masad explained that Replit evaluated models using both internal benchmarks and published public evaluations.
“I think ChatGPT is better now,” Masad said onstage. “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.”
In software development contexts, "vibe coding" describes workflows where builders use conversational prompts to generate, iterate, and deploy entire applications from scratch. According to Masad, OpenAI's Sol and Astra models demonstrated superior stability and execution when handling end-to-end application generation within Replit's testing frameworks.
Enterprise Expansion and Commercial Growth
Alongside model integrations, Masad shared metrics illustrating Replit's shifting commercial revenue mix. Although the company has not publicly disclosed Annual Recurring Revenue (ARR) totals, its customer composition has moved substantially toward organizational clients.
Enterprise contracts accounted for approximately 2% of company revenue in the previous year, expanding to roughly 15% in the current year. Business-to-business (B2B) agreements now represent more than 50% of Replit's total revenue. Backstage at Disrupt, Masad projected that enterprise and government clients will comprise more than half of the company's total revenue by 2027.
Architecting an 'Agent Habitat' for Safe AI Deployment
Addressing the broader operational requirements of autonomous programming tools, Masad outlined the necessity of robust security and execution environments for AI agents.
“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 said. “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.”
Masad highlighted that establishing reliable sandboxing, access controls, and network boundaries is essential before organizations can deploy autonomous coding agents safely across enterprise production pipelines.
Frequently Asked Questions
Why did Replit switch its default model provider from Anthropic to OpenAI?
According to Replit CEO Amjad Masad, the shift was determined by cost, partnership strength, and benchmark performance. Masad stated that OpenAI models, specifically Sol and Astra, consistently performed better on internal and public benchmarks for building vibe-coded applications from scratch.
Can developers still access Anthropic's Claude models on Replit?
Yes. Masad clarified that Replit remains model agnostic. While OpenAI serves as the default provider for Replit Agent, users retain the option to manually select Claude or other supported models.
Did Replit completely sever its partnership with Anthropic?
While Masad stated onstage at TechCrunch Disrupt that Replit was no longer working with Anthropic, an insider familiar with the arrangement rejected that claim and stated the partnership remains in place. Furthermore, Replit promoted Anthropic models, including Claude Fable 5, Claude Fable 5.1, and Opus 5.5, across official blog posts and social channels shortly before the conference.
What proportion of Replit's revenue comes from enterprise contracts?
Enterprise contracts have risen from roughly 2% of Replit's revenue last year to approximately 15% this year, with B2B contracts making up more than 50% of total revenue. Masad projected that enterprise and government contracts will represent more than half of revenue by 2027.
0 Comments