The Replit Agent default model is now OpenAI, not Anthropic’s Claude, according to Replit CEO Amjad Masad. Your projects keep working, and you can still pick Claude yourself. But if you never touch the model setting, the Agent now builds with OpenAI.
Here’s what Replit said, what’s still disputed, and what to check today.
Quick Facts
- Masad made the announcement onstage at TechCrunch Disrupt on October 15, 2026.
- OpenAI is the default for Replit Agent; Claude and other models remain available.
- Replit gave three reasons: cost, the partnership, and performance.
- The benchmark claims come from Replit, with no independent results.
That’s the short version. The details matter more, because the story changed a little after the talk.
What changed with the Replit Agent default model
Masad said Replit no longer sees Anthropic as its main partner and now considers OpenAI’s models stronger. He pointed to an internal benchmark and a public one, saying OpenAI does better at building apps from scratch. “I think ChatGPT is better now,” he said, as reported by TechCrunch via Yahoo Finance.
Replit announced its OpenAI partnership back in August. Masad later clarified that Replit “remains model agnostic,” so you can still choose Claude or another supported model, according to a DEV Community summary of the follow-up.
This default matters because most people never change it. TechCrunch noted that most customers stick with whatever the Agent starts with.
Why Replit says it made the switch
Masad gave three reasons: cost, the partnership, and performance. He called OpenAI “a better partner for now.” Replit was until recently a top-five Anthropic customer, so the move could cut its Anthropic bill a lot.
Cost is a live issue across AI coding tools. Model providers keep changing prices and speed tiers, like the Claude Opus 5.5 fast mode price jump we covered earlier. A platform that runs millions of Agent sessions feels those changes fast.
OpenAI is also pushing its own agent features, such as ChatGPT parallel agents for Pro users. A vibe coding platform picking its default partner fits that wider race.
What’s confirmed and what’s disputed
Not everything Masad said has held up cleanly. Here’s how the main claims stack up.
| Claim | Status |
|---|---|
| OpenAI is now the default model provider for Replit Agent | Confirmed by Masad onstage and in follow-up remarks |
| Claude and other models are still selectable | Confirmed by Masad after the event |
| Replit is “no longer working with” Anthropic | Disputed; a person familiar with the deal says the partnership continues |
| OpenAI models build apps from scratch better | Replit’s own claim; no independent results published |
| Effect on your credits or pricing | Not stated in any source I found |
The disputed part is worth a closer look. TechCrunch cited a person familiar with the relationship who said the Anthropic partnership is still in place. Replit’s own late-September blog post also named Claude models as running in Agent’s production environment.
Neither Anthropic nor Replit had issued a formal correction in the reports I found, and TechCrunch said it would update its story if Replit responded. That leaves a gap between a stage comment and the paper trail, which is why the wording matters here.
So treat “Replit dropped Anthropic” as an overstatement for now. “Replit changed its default to OpenAI” is the safe version.
How to check which model your Replit Agent uses
Open your project and look at the Agent settings for a model option. I couldn’t find Replit’s published steps for this change, so menu names may differ from what you see.
Users on any Replit plan who build with the Agent are the ones most likely to be affected, since the default applies unless you override it. Teams that share projects should agree on one model so everyone gets consistent results.
If you’re mid-project, don’t switch models halfway through without a reason. Different models make different choices about structure and libraries, and a mid-build swap can leave your code feeling inconsistent.
For a new project where quality matters, try the same prompt on both models and compare. It costs a little time and tells you more than any benchmark Replit or anyone else publishes.
Does the Replit Agent default model matter for your projects?
For hobby builds, probably not much. Both model families can produce a working app from a good prompt.
It matters more for teams. If your code review rules, security checks, or budget assumptions were tuned around Claude’s output, a new default can change style, speed and spend. Masad said enterprise work is now about 15% of Replit’s revenue, up from roughly 2% last year, so expect more attention on that audience.
Replit hasn’t said how the change affects credits or pricing. Watch your usage for a week before you assume it stays the same.
Frequently Asked Questions
Did Replit stop using Claude?
No, not entirely. Masad said users can still pick Claude and other models. A person familiar with the deal also told TechCrunch the Anthropic partnership continues.
Will my existing Replit projects break?
Nothing in the reports suggests they will. The change affects which model the Agent uses by default, not your saved code.
Is OpenAI actually better for building apps?
Replit says so, citing an internal benchmark and a public one. No independent results were published, so test it on your own prompts.
Will this change what I pay?
Unknown. None of the sources I found say how credits or pricing are affected.
My take: don’t panic and don’t ignore it. Check which model your Agent runs, run one side-by-side test on a real task, and keep the one that gives you cleaner code. The headline is about a business deal, but the choice that affects your project is a single setting.
Leave a Reply