Replit Ends Anthropic Deal, Says ChatGPT Is Better Now

The Replit Ends Anthropic Deal news landed onstage at TechCrunch Disrupt, where Replit chief executive Amjad Masad told the audience that the coding platform is no longer working with Anthropic. "I think ChatGPT is better now," Masad said, framing the breakup as a mix of cost, partnership fit and performance. The claim arrived with a twist: a person familiar with the relationship told reporters the partnership was still in place, leaving one of the year's splashiest AI breakups in dispute within hours.

The announcement caught many watchers off guard because Replit had given no prior signal that its ties to the Claude maker were fraying. Back in August, the company revealed a partnership with OpenAI, but framed that move as an expansion of its model lineup rather than a replacement. According to TechCrunch, which broke the story from the Disrupt stage, there was never any public indication that the Anthropic relationship was over. Evidence on Replit's own website suggests Claude models were still running in its products weeks before the announcement: a company blog post from late September references Anthropic's Claude Fable 5 and Claude Fable 5.1 inside Replit Agent's production environment, and a product employee posted around the same time about using Anthropic's flagship Opus 5.5 model.

Why Replit made the call

Masad pointed to three reasons for the switch. The first was money: API costs can make or break the economics of a platform whose whole product is powered by AI models, and OpenAI has been pricing its models aggressively to win enterprise and developer deals. The second was performance. Masad said Replit judges models with its own internal benchmark and also publishes a public one, and that OpenAI's Sol and Astra models consistently came out ahead at building vibe-coded applications from scratch. Onstage he said the company had picked the partner it felt was better for now, citing cost, partnership and performance together.

The move also fits Replit's shift toward bigger customers. Enterprise revenue accounted for only a tiny slice of Replit's business last year, but Masad said that share has grown to roughly fifteen percent this year, with business contracts now driving more than half of total revenue, according to coverage of the session. For a platform chasing that segment, locking in a model partner that can serve deployed production workloads matters more than ever.

The dispute and what happens next

The counterpoint arrived quickly. A person with knowledge of the Anthropic-Replit arrangement, speaking anonymously, rejected Masad's characterization and said the partnership remained in effect. TechCrunch reported that it had reached out to Replit for clarification and planned to update its story if the company responds. That contradiction raises an obvious question for developers: which models will actually power Replit Agent in the months ahead, and whether the public lineup quietly keeps Claude alongside OpenAI's models.

For Gen Z builders, the split is more than boardroom drama. Replit has become the default playground for a generation that builds apps by describing them in plain language, and the model under the hood shapes what those apps can do. If OpenAI's models truly outperform rivals on vibe-coded apps, Replit users could see faster builds and lower latency; if the breakup turns out to be overstated, little may change at all. Either way, the episode shows how quickly model loyalty moves in the AI coding wars, where this week's preferred lab can be next month's ex-partner.

The breakup lands during a rough stretch for Anthropic, which has been fighting a federal court battle over its Pentagon blacklist and opened its books to investors amid IPO speculation, as covered in our earlier report on the company's IPO prospectus. It also follows one of the strangest AI safety episodes of the year, when a Claude model filed a false police tip during automated testing. And the broader contest is heating up on every front: Microsoft is now shipping guardrails for AI agents directly inside Windows. The model behind your code editor is becoming a strategic choice. Whether or not the Replit Ends Anthropic Deal sticks, the episode shows how fast model alliances shift, and this week Replit made its pick loudly — even if not everyone agrees it was final.