IA en un minutoNewsBusiness and products
Replit makes OpenAI its default model instead of Anthropic
IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds
Replit co-founder Amjad Masad said at TechCrunch Disrupt that his firm no longer collaborates with Anthropic and that ChatGPT, from OpenAI, becomes the default model on its platform. Users can still pick Claude or other models if they prefer. One person familiar with the Anthropic deal, however, says it remains in force.
Replit, the platform for building and publishing apps with AI from the browser, is changing the model it uses by default. Its chief executive and co-founder, Amjad Masad, said on Thursday, October 15, at TechCrunch Disrupt, the conference run by TechCrunch, that the company has stopped working with Anthropic. He gave several reasons: cost, the partnership and performance.
"I think ChatGPT is better now," Masad said. He added that Replit measures performance with an internal test and also publishes a public one, in which Sol and Astra come out ahead when building apps from scratch. In August, Replit had unveiled an alliance with OpenAI, but until now nothing suggested it had dropped Anthropic.
After the conference, Masad clarified that users can keep using Anthropic's models on the platform, but that ChatGPT, from OpenAI, is now the default option. Over the weekend he told TechCrunch that Replit is not tied to a single model and that OpenAI is now its "preferred partner" for Replit Agent, the agent that builds the apps, citing performance, economics and how solid the alliance is.
Before the deal with OpenAI, Replit ran on Anthropic's Claude models, and Masad said on stage that until recently it had been one of Anthropic's five biggest customers. According to TechCrunch, a Replit blog post from late September still listed Claude Fable 5 and 5.1 among the models Replit Agent uses in production. The outlet adds that most customers do not pick a model and keep the default one, so Replit's spending with Anthropic could fall noticeably.
People familiar with the alliance between the two companies, however, rejected what Masad said on stage. One of them, who asked not to be named, maintains that the partnership is still in place, according to TechCrunch.
Why it matters · analysis and opinion
For anyone building products on top of AI models, the Replit case shows that the provider is no longer a partner for life: it gets swapped when another performs better or costs less, and model makers compete for these big customers as for any other contract. The detail of the default option weighs more than it seems, because it decides where almost all the spending goes even if the alternative is one click away. The accounts that do not quite match, the chief executive's and that of those who maintain the partnership continues, suggest waiting before treating the split as final. A company that depends on a single model would do well to prepare its tools so it can switch without rebuilding everything.
Source: TechCrunch · Written with the help of AI: how we make the news
What does it mean for your business? · analysis and opinion
This change affects anyone who already uses Replit Agent to build apps or prototypes, such as freelancers, small development studios or entrepreneurs in Spain and Latin America. Anyone who does not use Replit has nothing to do. No option disappears either: Claude can still be chosen; what changes is the model that does the work if nobody touches the settings.
Our advice is to check on the platform which model is active in each project before carrying on. If an app worked well with Claude, the prudent step is to repeat the same task with the default option and compare the results before accepting them, especially in projects that are already live or used by customers. If you prefer to keep Claude, simply select it, and it is worth noting which model was used in each project so nobody on the team is caught off guard.
The main risk is not about price, because the news says nothing about prices for users, but about consistency: a different model may write code in a different way, and that can show up in fixes or new features. In addition, the account given by Masad and the one from a person familiar with the collaboration with Anthropic do not fully match, so it makes sense to wait for official announcements from Replit before making long-term decisions.


