Replit is no longer working with Anthropic, the company's CEO and co-founder Amjad Masad said Thursday while onstage at TechCrunch Disrupt — a striking shift for the AI coding platform that long ran on Claude models and has become a favorite of startup founders and developers.
Masad said the decision was made for a number of reasons, including cost. "I think ChatGPT is better now," he said, adding that the company uses an internal benchmark to help assess metrics like performance. "We publish a public benchmark as well, and consistently, Sol and Astra perform better on building vibe-coded applications from scratch." For more context on this story, see our AI industry coverage.
Cost, Partnership, and Performance
Replit announced a partnership with OpenAI in August, but there had been no indication that it was no longer working with Anthropic. Masad framed the switch as the result of deliberate evaluation rather than a sudden falling-out.
"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," he said, referring to OpenAI.
Replit sells a cloud-based software development platform where users can develop and deploy apps directly from inside their browser. Under its partnership with Anthropic, Replit used the AI lab's Claude models to run its platform — the models powered Replit Agent, the feature that generates and modifies whole applications from natural-language prompts, and the arrangement was said to have helped the company grow its revenue significantly.
That positioning explains why the onstage comments landed hard. If a platform whose core promise is "build and deploy software from your browser" swaps out the model doing the actual software generation, customers and rivals will read it as a verdict on which lab's models are currently best at agentic coding.
Insiders Say the Anthropic Partnership Is Still in Place
There is a complication: Masad's comments were rejected by people familiar with the partnership. One person familiar with the arrangement, who asked not to be named, told TechCrunch that the partnership was still in place.
There is also documentary evidence that Replit was using Anthropic products as recently as a few weeks ago. A blog post in late September cites Anthropic's Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent's production environment. Around the same time, one of Replit's product employees posted on X about using Anthropic's flagship Opus 5.5 model and promoted its capabilities.
TechCrunch reached back out to Replit for clarification and said it would update the story if it hears back. Until then, the picture is genuinely ambiguous: the CEO says the relationship is over, while people familiar with it and Replit's own recent materials suggest Claude may still be running in production.
Enterprise Is Now a Meaningful Slice of Revenue
Masad shared growth figures while onstage, and they help explain the commercial logic of the switch. Enterprise was about 2% of Replit's revenue last year, he said; this year, enterprise contracts account for roughly 15% of revenue, and more than 50% of overall revenue is business-to-business.
"We're not sharing ARR numbers just yet," he added. The shift toward enterprise and B2B customers changes the calculus on model costs: at scale, the price per token across millions of generated apps is a material line item, and Masad explicitly cited cost as one of the deciding factors.
Masad also addressed AI safety measures during the appearance, telling the audience that safety requirements for AI-built software deserve the same seriousness that human engineering practices do.
What the Switch Says About the Model Market
Two things stand out from the Disrupt conversation. First, frontier model switching has become operationally routine for AI application companies — Replit moved the engine under a production product and discussed it casually onstage, the way startups once discussed changing cloud providers.
Second, evaluation benchmarks are now the battlefield. Masad pointed to both an internal benchmark and a public one to justify the decision, and his claim that OpenAI's Sol and Astra models "perform better on building vibe-coded applications from scratch" is exactly the kind of head-to-head comparison that model vendors now live and die by.
For Anthropic, losing — or demoting — a marquee developer-platform customer is a reputational hit even if the revenue relationship continues in some form. For OpenAI, it is a proof point in its push to win the AI coding market. And for developers watching the platform wars, the lesson is that no model partnership is permanent: the benchmark leaderboard, and the invoice, decide.
Replit did not respond to follow-up questions from TechCrunch before publication, and Anthropic has not publicly commented on Masad's remarks.
---
Get the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
