Replit CEO says the company is no longer working with Anthropic

Replit CEO and co-founder Amjad Masad said the AI coding platform is no longer working with Anthropic, according to TechCrunch, marking a notable shift for a browser-based development tool used by founders and developers. The move matters because AI coding agents increasingly depend on underlying model providers, and those choices can affect cost, behavior and trust for teams building apps with AI.

The situation is not entirely settled. TechCrunch reported that people familiar with the Anthropic-Replit partnership disputed Masad’s comments, and the outlet said it had asked Replit for clarification.

What Masad said changed

Speaking onstage at TechCrunch Disrupt, Masad said Replit had moved away from Anthropic for several reasons, including cost. Replit announced a partnership with OpenAI in August, but TechCrunch noted there had been no earlier sign that the Anthropic relationship had ended.

Masad said, “I think ChatGPT is better now,” and pointed to Replit’s internal benchmark for assessing areas such as performance. He also said the company publishes a public benchmark, and that Replit’s Sol and Astra perform better at building “vibe-coded applications from scratch,” according to TechCrunch.

For users, the practical takeaway is that the model behind an AI coding product is not just a back-end detail. If a tool changes its preferred provider, generated code, debugging behavior, latency and failure modes may feel different, even if the product name stays the same.

Replit offers a cloud-based development environment where people can build and deploy apps from the browser. Under its Anthropic partnership, TechCrunch reported, Replit used Anthropic’s Claude models to run its platform, and that partnership was said to have helped the company grow revenue significantly.

The Anthropic picture is disputed

TechCrunch also reported that insiders familiar with the relationship pushed back on Masad’s characterization. One person familiar with the matter, speaking anonymously to the outlet, said the partnership was still in place.

There are also signs that Anthropic models were in use recently. TechCrunch pointed to Replit’s website, blog posts and social media posts as evidence that the company had been using Anthropic products as recently as a few weeks before the Disrupt appearance. One late-September blog post cited Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment, according to the report. Around the same period, a Replit product employee posted on X about using Anthropic’s Opus 5.5 model.

That ambiguity matters for teams choosing an AI coding stack. If you are building client work or internal tools with Claude, ChatGPT or a packaged coding environment, it is worth asking whether the vendor discloses which models are used, whether that can change, and how changes are communicated. For a broader comparison of major assistants, see our ChatGPT vs Claude vs Google Gemini buying guide.

Why model partnerships matter for AI app builders

Replit sits in a crowded AI coding market alongside tools built for assisted editing, browser-based app creation and agent-style software development. The underlying model provider is only one part of the product, but it can shape the experience in important ways.

For freelancers, marketers and small businesses using AI coding agents to create landing pages, internal dashboards or lightweight apps, the key risks are practical rather than theoretical:

• A model switch can change how an agent interprets prompts or fixes bugs.
• Costs can influence which model a platform routes work to.
• Benchmarks may not match your own project type, stack or quality bar.
• Safety, sandboxing and deployment controls become more important as agents do more work.

Masad framed Replit’s decision as a combination of cost, partnership and performance. He said Replit considered OpenAI a better partner “for now,” according to TechCrunch. That phrasing leaves room for future changes, and it underlines how fluid the AI infrastructure layer remains.

Enterprise growth and safety are now part of the story

Masad also discussed Replit’s business mix at Disrupt. He said Replit is not yet sharing annual recurring revenue numbers, but that enterprise customers have become a larger part of revenue. According to Masad, enterprise was about 2% of revenue last year and is about 15% this year, while more than half of revenue is now business-to-business.

That shift helps explain why model choice, reliability and safety controls are becoming more visible. Larger customers tend to care about governance, security and consistency, especially when AI agents can create and deploy software.

Masad said agents need a proper “habitat” to do useful work safely, including orchestration, error recovery, safety, security, sandboxing, networking, identity and authorization. In other words, the model is only one layer. For businesses experimenting with AI-built apps, the surrounding controls may be just as important as whether the agent is powered by OpenAI, Anthropic or another provider.

For now, the clearest fact is that Replit’s CEO says the company is no longer working with Anthropic, while TechCrunch reports that people familiar with the partnership dispute that account. Until Replit provides more detail, users should treat the platform’s model stack as something to verify rather than assume.