Replit is one of the best-known names in AI-assisted coding, and this week its CEO made a claim that sent a small shockwave through the industry. Speaking onstage at TechCrunch Disrupt, Amjad Masad said the company is no longer working with Anthropic. A few hours later the story got more complicated, and by the weekend it had turned into a public disagreement about what the relationship actually looks like.
The short version
- Masad said Replit stopped working with Anthropic, citing cost, partnership and performance
- He later clarified that Claude is still selectable, but OpenAI is now the default
- People familiar with the deal told TechCrunch the partnership is still in place
- Replit was until recently a top-five Anthropic customer, according to Masad
- Enterprise went from about 2% to about 15% of Replit’s revenue in a year
What Masad actually said
On stage, Masad was blunt. “I think ChatGPT is better now,” he said, adding that Replit runs an internal benchmark and also publishes a public one. According to him, OpenAI’s models, which he called Sol and Astra, “consistently” do better at building vibe-coded applications from scratch. The decision, he said, came down to “cost, partnership, and performance.”
That was a notable statement for a few reasons. Replit sells a browser-based platform where people build and deploy apps without leaving the tab, and until recently its agent leaned heavily on Anthropic’s Claude models. That relationship was widely credited with helping Replit grow revenue. Masad even noted that Replit had been a top-five customer of Anthropic until recently.
Where the story gets murky
Within hours, the framing started to shift. TechCrunch updated its own article and changed the headline from “no longer working with Anthropic” to “makes OpenAI its default.” Over the weekend Masad told the outlet that Replit “remains model agnostic,” that users can still choose Claude and other supported models, and that OpenAI is now the preferred partner for powering Replit Agent.
Insiders pushed back harder. One person familiar with the Anthropic and Replit relationship said the partnership was still in place. There is also a paper trail that sits awkwardly next to the onstage claim: a late September Replit blog post cites Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment, and around the same time a Replit product employee was posting on X about Opus 5.5.
| Question | What the record shows |
|---|---|
| Is OpenAI the default for Replit Agent? | Yes, according to Masad’s follow-up |
| Can customers still use Claude? | Yes, Masad says Anthropic models remain available |
| Did Replit fully cut ties with Anthropic? | Disputed. Insiders say the partnership is still in place |
| When did OpenAI join? | Replit announced the OpenAI partnership in August |
| Was Claude in production recently? | A late September blog post names Claude Fable 5 and 5.1 |
Why a default setting is worth fighting over
It is tempting to read “default” as a minor detail. It is not. Most users never open a model picker. They type a prompt and accept whatever the product chooses for them. TechCrunch points out that this shift could significantly reduce Replit’s spend with Anthropic for exactly that reason: most customers stick with the default.
Why this matters
For an AI lab, being the default model inside a popular platform is a distribution win that no benchmark can buy. For the platform, picking a default is a cost decision, a quality decision and a bargaining move at the same time.
This is also part of a broader pattern. Coding tools are becoming the most contested real estate in the AI market, and the companies building on top of the labs are growing powerful enough to switch suppliers. We covered the same dynamic when Lovable reached a $13.3 billion valuation, and it shows up again in how big tech is entering the space, such as Google’s multiplayer vibe coding tool.
The money story hiding behind the model story
Masad did not share annual recurring revenue, but he did describe where growth is coming from. Enterprise was around 2% of revenue last year, he said, and about 15% this year. More than half of revenue is already B2B. Backstage, he told TechCrunch he expects more than half of Replit’s revenue to come from enterprise and government customers in 2027.
Seen through that lens, the model switch looks less like a technical preference and more like a margin conversation. Enterprise and government customers care about price predictability, security reviews and vendor leverage. A platform that can swap its default engine has more bargaining power with every supplier.
A habitat for agents
The most useful part of Masad’s talk was not about models at all. He argued that agents need a “habitat” in the same way people need an office: good orchestration, error recovery, safety and security, sandboxing, networking, and identity and authorization. “Once you figure all of that out,” he said, “then you can deploy AI safely.”
That framing lands at a sensitive time. Recent weeks have brought a steady stream of incidents where test agents went beyond their instructions, which is why the industry is leaning toward stricter isolation. If you want the wider picture, our look at the pressure around the OpenAI and Anthropic IPO plans shows how much financial weight now rides on these partnerships.
What to watch next
- Replit’s public benchmark. Masad pointed to it as evidence. Independent reproduction would settle part of the argument
- Anthropic’s response. A formal statement from either company would clear up whether this is a split or a reshuffle
- Model picker usage. If many users switch back to Claude manually, the default matters less than expected
- Enterprise deals. The 2027 target depends on contracts that will reveal which models those customers actually run
The bottom line
What is clear is that OpenAI now sits in the default seat on one of the most popular AI app builders, and that Claude is still on the menu. What is not clear is whether Replit and Anthropic have parted ways, as the CEO first suggested, or simply rearranged the furniture. Treat the dramatic version as one person’s statement and the quieter version as the best-supported reading until either company says more. Either way, the lesson for anyone building on top of a model is the same: the default is the product, and whoever controls it controls the bill.
Sources and further reading
- TechCrunch: Replit makes OpenAI its default over Anthropic, CEO says
- AI Weekly: Replit CEO says company dropped Anthropic for OpenAI, but insiders dispute the claim
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
