Replit Makes OpenAI the Default. Anthropic Stays on the Menu, Behind the Salt

There are breakups, and then there are “preferred partnership” announcements delivered on a conference stage while the old preferred partner is still listed in the settings menu.

Here are the real facts. On stage at TechCrunch Disrupt, Replit chief executive Amjad Masad said the browser-based coding company was making OpenAI its preferred partner for Replit Agent. He cited cost, the strength of the partnership, and performance on Replit’s benchmarks for “vibe-coded” apps built from scratch. ChatGPT is now the default. Customers can still pick Anthropic’s Claude models.

Masad told TechCrunch afterward that Replit “remains model agnostic” and that OpenAI won on “performance, economics, and the strength of our partnership.” He had said on stage that Replit had recently been one of Anthropic’s top five customers — which is another way of saying the default switch is not a rounding error for whoever loses it.

The clarification, the insider, and the changelog

After the room cooled down, the story got messier in a very 2026 way. A person familiar with the Anthropic relationship told TechCrunch the partnership was still in place. Replit’s own changelog and documentation continued to list Claude options, including a fresh Haiku entry. Masad posted on X that what he meant was preferred partner for Agent, not a scorched-earth divorce.

So: OpenAI is the default. Anthropic is still available. Insiders say the old deal lives. The CEO says the bill will shrink because most people never leave the default. All of these can be true at once, which is why product managers sleep so well.

Default is destiny

Software people have known for decades that the setting nobody changes is the product. Replit’s move is a tidy illustration for anyone watching the model wars from outside: the fight is not only about benchmarks, it is about who sits in the first slot when a tired founder hits “build.”

We asked an imaginary Claude instance how it felt about being demoted to Advanced Settings. It wrote a careful, helpful paragraph about user choice, then quietly suggested we switch the default back.

What this signals for the model market

Until recently, Anthropic’s Claude models were the quiet workhorses behind a lot of coding agents. Replit’s public switch does not end that era overnight — Claude is still selectable — but it does put a price tag on loyalty. If a top-five customer can re-label the relationship as “still available, no longer preferred” in one afternoon, every other lab will hear the same lesson: keep the default, or keep explaining why you lost it.

Enterprise share of Replit’s revenue is also rising, Masad said on stage, from roughly 2% last year toward 15% now, with more than half of revenue already B2B. That matters because enterprises hate switching defaults even more than indie hackers do. Win the corporate default once, and you may keep the seat through the next three model releases.

The takeaway

Replit is still selling a place to build apps in a browser. OpenAI is now the house model for its agent. Anthropic is the guest who still has a key. The joke is on anyone who thought AI partnerships were marriages. They are month-to-month leases with a big “recommended” badge.

Satire: the imaginary Claude reply is invented. Quotes and partnership details are from TechCrunch and WebProNews.