Replit just made it cheaper to build software without hiring a developer, and the reason why says a lot about where AI pricing is headed.
Replit is a platform where you describe what you want in plain English and its AI assistant, called Agent, writes and runs the software for you. No coding needed. The company calls this Free Mode, a new setting that lets subscribers do up to 30 times more everyday work, like quick chats, planning, and small tasks, without spending down their monthly usage allowance.
The generosity has a specific source. Free Mode runs only on OpenAI's GPT-5.6 Luna, a fast, low-cost model that OpenAI cut in price by 80 percent this past July, down to 20 cents per million words of input. That is a fraction of what rival models from Anthropic charge. Replit is passing that savings straight to users instead of pocketing it, which is smart: cheap access keeps people building on Replit rather than a rival.
It is worth being clear about what Free Mode actually covers. It handles the light stuff: chatting with the assistant, planning a project, small fixes. The moment a task gets bigger or more complex, Replit nudges you toward Power Mode or Max Mode, which still draw from your paid credits. So the 30 dollars a month plan is not an all-you-can-build buffet. It is a much bigger free lane for the everyday work, with the real cost showing up once you try to ship something serious.
The bigger story here is Replit's growth, and what it says about demand from non-technical builders. The company's revenue went from about 2.8 million dollars in 2024 to a run rate approaching 1 billion dollars by the end of this year, according to statements from CEO Amjad Masad. Its valuation tripled to 9 billion dollars within six months. That growth is not coming from professional software teams. It is coming from people who have never written a line of code, using Replit to build businesses, internal tools, and side projects.
This also deepens the tie between Replit and OpenAI. Choosing to run Free Mode exclusively on OpenAI's model, rather than mixing in Anthropic's Claude the way Replit has in the past, signals a closer working relationship between the two companies. For OpenAI, Replit is a distribution channel that puts its cheapest model directly in front of millions of builders who will never touch an API. For Replit, it locks in a supplier with strong incentive to keep costs falling.
There is friction elsewhere worth watching. Replit has an active dispute with Apple, which has blocked updates to Replit's iPhone app over concerns about code changing after the app is installed. That is a preview of a coming headache for any tool that builds and updates software on the fly: app store rules were written for software that does not rewrite itself, and AI tools increasingly do.
For a business thinking about using tools like this, the takeaway is simple. Testing a small internal app, a form, a tracker, a simple calculator, now costs close to nothing on a plan you are probably already paying for. Building something you plan to run your business on, or ship to customers, still costs real money and still needs someone who understands what good software looks like before you trust it with your data.