Replit CEO Says Company No Longer Works With Anthropic; Insiders Dispute That It Has Ended

By The News Desk, Previously

October 17, 2026

There are cleaner ways to end a business relationship than announcing it from a conference stage, but subtlety has never really been the AI industry’s house style. On Thursday, October 16, onstage at TechCrunch Disrupt, Replit CEO and co-founder Amjad Masad said his company is no longer working with Anthropic — the AI lab whose Claude models powered Replit’s rise — according to TechCrunch.

Asked whether the OpenAI partnership Replit announced in August meant the Anthropic relationship was over, Masad answered, “Yeah. I mean, yeah,” according to Unite.AI’s account of the session. He framed the decision around three things: “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,” according to TechCrunch.

“I Think ChatGPT Is Better Now”

The performance claim is the part Masad leaned on hardest. “I think ChatGPT is better now,” he said, according to TechCrunch, pointing to an internal benchmark Replit uses to assess models — an internal benchmark Unite.AI identified as VBench — and to a public benchmark the company publishes. On both, he said, OpenAI’s Sol and Astra models “consistently” perform better at building vibe-coded applications from scratch, according to TechCrunch. That specific task matters more than a generic leaderboard: Replit’s whole product is an AI agent that builds apps from a description inside a browser-based development platform, so the model that writes the best first draft of an app is, functionally, the product.

Masad also put a number on what Anthropic is supposedly losing. Replit had been, until recently, one of Anthropic’s top five customers, he said onstage, according to TechCrunch. Before the OpenAI deal, Replit relied on Anthropic’s Claude models to run its platform, and that partnership was said to have helped the company grow its revenue significantly, according to TechCrunch. Losing a top-five customer — if that is what happened — is not a rounding error for any AI lab, whatever the press releases say.

The Other Side: “Still in Place”

Here is where the story gets genuinely interesting, because the breakup may not be a breakup. Masad’s comments were rejected by insiders familiar with the partnership between Anthropic and Replit, according to TechCrunch: one person familiar with it, who asked not to be named, said the partnership was still in place. Technopolis reported the same dispute, summarizing that people familiar with the partnership say it is still in place even as Masad says Claude is out.

The paper trail leans, at minimum, toward “it’s complicated.” A Replit blog post in late September cited Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment, and around the same time a Replit product employee posted on X about using Anthropic’s flagship Opus 5.5 and promoting its capabilities, according to TechCrunch. Companies do not usually advertise a supplier’s newest models in production three weeks before firing that supplier — unless the firing was more improvised, or more narrow, than the headline version.

The Clarification That Reframes the Headline

After Disrupt, Masad himself supplied the narrower version. He clarified that customers can still access Anthropic models on the platform, but that OpenAI’s ChatGPT is now the default — a shift TechCrunch noted could significantly reduce Replit’s spend with Anthropic, since most customers stick with the default rather than choosing a model themselves. “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.”

Read together, the fairest summary is this: Replit has demoted Anthropic from the engine room to the options menu, its CEO described that demotion onstage in the language of a divorce, and people close to the partnership insist the legal relationship itself is still standing. All three of those things can be true at once, and in platform economics the default is the ballgame anyway — whoever holds the default holds the revenue, whether or not the contract has technically ended.

Why Replit Can Afford to Play Hardball

Masad also talked growth onstage, and it explains his leverage. Enterprise was about 2% of Replit’s revenue last year and is roughly 15% this year, with more than half of revenue now coming from business customers, and he expects more than half of revenue to come from enterprise and government customers in 2027, according to TechCrunch. A company growing like that renegotiates its supplier deals from strength. He also made the case that Replit is not merely a wrapper waiting to be absorbed by the labs: its agents can spin up databases and deploy into government and enterprise environments, and, as he put it, agents need a proper habitat — “good orchestration, good error recovery, good safety and security and sandboxing,” according to TechCrunch.

TechCrunch said it reached back out to Replit for clarification and would update its story. Until somebody produces a contract, the honest headline is the messy one: the CEO says it’s over, the insiders say it isn’t, and the default setting — the only vote that spends money — has already changed. Follow the rest of our coverage in Previously Tech and News.