Replit Is Switching the AI Behind Its Coding Tool

Replit is no longer working with Anthropic, co-founder and CEO Amjad Masad said onstage at TechCrunch Disrupt. TechCrunch
Masad named cost, partnership and performance as the reasons. Before its August deal with OpenAI, Replit relied on Anthropic's Claude models to run its platform. Until recently, Replit was one of Anthropic's five largest customers.
The switch is now live. OpenAI's ChatGPT is now the default, the normal option, on Replit. OpenAI is now Replit's preferred partner for Replit Agent, the assistant that does multi-step coding tasks. Like a restaurant with a house dish, the normal choice is set but other dishes stay on the menu. Customers can still use Claude or other supported models. Replit says it remains model agnostic, which means users can choose.
The news follows several launches. On August 18, 2026, Replit announced Free Mode to expand what is possible with AI, stating that with OpenAI, Replit helps users accomplish more every day. On September 10, 2026, Replit said its Databricks integration is generally available with native Lakebase support. On September 25, 2026, Replit announced it is acquiring Atta to add interactive charts in chat.
More announcements followed. On September 29, 2026, Replit published work on agent harness design under the title Free the models: Harness design at the frontier, stating Replit Agent lets the model choose its subagents' tier and effort. On October 13, 2026, Replit introduced Replit Solutions Partners to help organizations put code to work. On October 16, 2026, Replit announced Bring Your Own Cloud, offering enterprise isolation and control.
The broader context here is the difference between choice and default. Users see a simple picker. Underneath is support software that manages tools, checks, retries and costs. That setup is tuned for one preferred model. Changing it takes work. It changes usage and cost, speed on long tasks, and how the main model shares work with smaller helpers.
In my view, the business and enterprise parts matter as much as the model name. Running these assistants for many users costs a lot. A top-five customer leaving gives both sides bargaining power. Lower costs can pay for free plans and longer runs. Losing that coding work also removes many real examples labs use to improve planning and tool use. Bring Your Own Cloud splits controls so customers keep isolation while Replit runs the assistant on top. Paired with the Databricks link and a partner program, the idea is to start with developers and then move into company data systems. Buyers should compare the default with other choices, check where data lives under BYOC, and ask how checks track quality as the preferred model changes. If cost and speed hold, users get more assistant time for each dollar. That limit has slowed these tools more than code quality, and cheaper runs allow longer plans, more tool uses and more tries before a person reviews.


