Replit switches its default model to OpenAI: pin yours

Replit just changed the default model under its coding agent. Most users never touch the default, so most Replit apps just got a new brain. CEO Amjad Masad said on stage at TechCrunch Disrupt that OpenAI's models are now the default for Replit Agent, replacing Anthropic's Claude, per TechCrunch. Claude is still available. You have to pick it.

What actually happened

Per TechCrunch's report from Disrupt:

• The switch. OpenAI models are now the default for Replit Agent. Customers can still select Anthropic's Claude models.
• The reasons. Masad cited "cost, partnership, and performance." He said Replit's internal and public benchmarks show OpenAI's models doing better at building apps from scratch.
• The official line. In a follow-up statement, Masad said Replit "remains model agnostic" and that OpenAI is now its "preferred partner" for Replit Agent. Replit announced an OpenAI partnership in August.
• The scale. Until recently, Replit was one of Anthropic's top five customers, TechCrunch reports. Because most users keep the default, the change could cut Replit's Anthropic spend sharply.
• The dispute. A person familiar with the partnership told TechCrunch the Anthropic arrangement is still in place, which conflicts with Masad's account. TechCrunch did not include a statement from Anthropic.

Why it matters for your business: your default model is a vendor decision

A coding platform's default model is not a setting. It is a procurement decision someone else makes for you, for their reasons: margin, partnership terms, their benchmarks. None of those are your app.

Late in September, Replit was still promoting Claude in its production stack. Weeks later, the default changed. That is how fast it moves.

If an agent wrote or runs part of your business on Replit, or any platform with a default model, here is what we would do this week:

• Pin the model. If your workflow depends on how a model behaves, pick it explicitly. Do not inherit the default.
• Run your own evals. Masad's benchmarks measure building apps from scratch. Your job is probably fixing and extending an app that already exists. Test the new default on ten real tasks from your backlog before you trust it.
• Check the bill. A different default model means different token prices and different token counts. Look at your usage after the switch.
• Keep the code portable. Your app should not care which model wrote it. Keep it in your own Git repo with plain, standard dependencies.

Sources: TechCrunch: Replit makes OpenAI its default over Anthropic, CEO says.