Replit knocks Claude off the top spot: OpenAI becomes the default model, while Anthropic insists the partnership still stands

Until recently Replit was among the five biggest customers of Anthropic. This week CEO and co-founder Amjad Masad walked on stage and said the company no longer works with Anthropic. "I think ChatGPT is better now," he said.

Masad listed three reasons: price, partnership and performance. According to Replit's internal and public tests, OpenAI's models are better at building apps from scratch. The OpenAI partnership was announced back in August, but nobody said at the time that Anthropic was on its way out.

Then came the softening. After the event Masad clarified that users can still choose Claude, but OpenAI is now the default model. "Replit remains model-neutral," he said. The difference is bigger than it sounds: most users never change what's been pre-selected for them. Replit's spending with Anthropic can shrink considerably without anyone formally tearing up a contract.

From the other side, the story is different. A source familiar with the relationship, who asked to remain anonymous, insists the partnership is still in place. And at the end of September a Replit blog post listed Claude Fable 5 and Fable 5.1 as part of their agent's production stack, while a company employee was publicly praising Opus 5.5. So what changed in a few weeks - the models, or the price?

Masad talked money too, but selectively. Enterprise deals brought in around 2 percent of revenue last year and around 15 percent this year, while more than half of revenue comes from other companies. Next year he expects enterprise and government customers to bring in more than half of revenue. Annual revenue, he says, they're not disclosing for now.

The lesson is as old as the market itself: in the AI model race, loyalty lasts until the next invoice. Today it's OpenAI, tomorrow it could be someone else - and a Replit user might not even notice that something changed under the hood.