TL;DR: Replit has made OpenAI its default AI model provider for Replit Agent, pushing Anthropic’s Claude out of the default slot. CEO Amjad Masad announced the switch at TechCrunch Disrupt on Thursday, October 15, citing cost, performance and partnership. Claude models are still selectable, and people close to the Anthropic relationship say that partnership is still in place. For most Replit users the practical change is simple: new Agent sessions run on OpenAI models unless you pick something else.
The news that Replit made OpenAI its default over Anthropic is a big signal in the AI coding market. Replit was, by Masad’s own account, one of Anthropic’s top five customers until recently. Because most people never touch the model picker, whoever holds the default slot on a platform this size gets most of the usage and most of the spend.
What Replit’s CEO actually said
Speaking onstage at Disrupt, Masad first described the change as Replit no longer working with Anthropic, according to TechCrunch. He framed it as a call based on quality and price:
“I think ChatGPT is better now. We publish a public benchmark as well, and consistently, Sol and Astra perform better on building vibe-coded applications from scratch. It was a decision based on both cost, partnership, and performance.”
“Sol” and “Astra” refer to OpenAI’s GPT-6 family models. The benchmark comparison is Replit’s own and hasn’t been independently checked.
After the event, Masad softened the message. He told TechCrunch that “Replit remains model agnostic,” that users can still choose Claude or other supported models, and that “OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.” WebProNews reports that he repeated the point on X on October 17, adding in a reply that “OpenAI models are generally more affordable.”
Anthropic isn’t fully gone
The “no longer working with Anthropic” line didn’t hold up well. TechCrunch reported that people familiar with the relationship rejected Masad’s onstage comments, and one of them said the partnership was still in place. There’s also a public paper trail:
- A late-September Replit blog post listed Anthropic’s Claude Fable 5 and Claude Fable 5.1 among the models running in Replit Agent’s production environment, per TechCrunch.
- Replit’s October 2 changelog added models from both companies: GPT-6.1 Sol in Max Mode and Claude Sonnet 5.5 in Power Mode, according to WebProNews.
- A day after the Disrupt talk, Replit’s October 16 changelog said Core and Pro users can now pick Claude Haiku 5.5 from the model selector in Power Mode, WebProNews reported.
So the accurate summary is that OpenAI won the default slot and the “preferred partner” title, while Claude stays on the menu.
What changes for Replit users
| Question | Answer, based on what Replit has said |
|---|---|
| Which model runs Replit Agent by default? | OpenAI’s models (Masad called ChatGPT the new default) |
| Can I still use Claude? | Yes. Claude models remain available in the model selector |
| Is Anthropic still a partner? | Disputed. Masad’s onstage comments said no; a person familiar with the deal told TechCrunch it is still in place |
| Why the switch? | Cost, performance on Replit’s benchmarks, and the strength of the OpenAI partnership, per Masad |
| When did the OpenAI partnership start? | Replit announced it in August |
If you liked Claude’s output better
If your projects behaved better on Claude, check the model selector before your next Agent session rather than assuming nothing changed. Replit’s changelogs show Claude options in Power Mode, and the exact choices depend on your plan. It’s also worth re-testing a small task on both models, because a different default can change code style, how often the agent asks questions, and how many steps it takes to finish.
Why this matters beyond Replit
AI coding tools are some of the biggest buyers of model capacity, and they don’t have to be loyal. Replit’s move shows how quickly a major customer can shift spend when one lab’s models get cheaper or score better on the tasks it cares about. TechCrunch noted that because most customers stick with the default rather than picking a model, the switch could significantly reduce Replit’s spending with Anthropic.
It also fits a wider trend we’ve been tracking on the site: platforms are turning into “agent habitats” that can swap the model underneath. We’ve covered how ChatGPT can now act on your Chrome tabs, how Perplexity open-sourced its agent sandbox, and how open-weight models such as Reflection’s Beam are aimed squarely at coding agents. Masad made the same argument at Disrupt:
“Agents need a habitat for them to exist and do good work. So that habitat has good orchestration, good error recovery, good safety and security and sandboxing, good networking, good identity and authorization.”
If you’re new to the idea, our explainer on agentic AI workflows covers the patterns these coding agents use.
Replit’s business is tilting toward enterprise
Masad also shared some growth details, without disclosing annual recurring revenue:
- Enterprise contracts were about 2% of revenue last year and about 15% this year, he said.
- More than 50% of Replit’s revenue is now business-to-business.
- Backstage, he told TechCrunch he expects more than half of revenue to come from enterprise and government customers in 2027.
That shift helps explain the focus on cost. Enterprise buyers care about predictable per-task pricing, and Masad said publicly that OpenAI’s models are generally more affordable for Replit. Those are Replit’s own claims about its own costs.
What to watch next
- Anthropic’s response. So far the pushback has come from unnamed people familiar with the deal, not an on-record statement.
- Independent results. Replit’s benchmark favors OpenAI’s models for building apps from scratch. Third-party tests on real projects will show whether users notice a difference.
- Other coding platforms. If price per task keeps falling, more tools may rotate their defaults the way Replit just did.
For everyday builders, the takeaway is practical: the model behind your coding agent can change without you changing anything. Check the selector, compare results on a real task, and pick what works for your project. If you’re also weighing a personal subscription, our breakdown of whether ChatGPT Plus is worth it covers what OpenAI’s paid tiers include.
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
