Replit has launched a new default mode for its paying subscribers that leans on one of OpenAI's cheapest models to keep everyday coding tasks from burning through a user's token budget, according to Fortune. The feature, called Free Mode despite requiring a paid plan, is now the default for subscribers on Replit's $20 a month Core tier and $100 a month Pro tier.
Free Mode runs on OpenAI's GPT-5.6 Luna model and handles routine work like chatting, brainstorming, and simple coding assignments without touching a user's token allowance. When a task turns out to need more horsepower, the system automatically reroutes it to a more capable model, then switches back to Free Mode once that heavier task is done.
The partnership only became financially viable after OpenAI cut Luna's API costs by 80 percent on July 30. Thibault Sottiaux, OpenAI's head of core products, said the reduction came from running the model more efficiently rather than from any sudden influx of new compute supply.
Michele Catasta, Replit's AI president, called the strategy radical, arguing that a lot of tasks simply do not require frontier level intelligence to complete well. OpenAI CEO Sam Altman framed the bigger picture in more sweeping terms, saying that if the industry can reach a point where anyone with internet access can build a real product, it could kick off something like a renaissance level entrepreneurial boom.
The move also reflects pressure both companies are under from two directions: enterprise customers increasingly skeptical about whether AI spending is paying off, and cheaper Chinese AI models eating into the market for anyone charging a premium. Making a capable model available at a much lower cost is one way to answer both concerns at once, betting that broader access to good enough AI will outperform charging steeply for the best available model.