Token costs drop out of Replit's new Free Mode

Replit has opened its app-building tools to users who are not paying for tokens, through a new Free Mode running on OpenAI's GPT-5.6 Luna. OpenAI says the mode is meant to let anyone turn an idea into working software without weighing token costs.

Metered token spend has been the main friction in agentic coding tools: the further a build goes, the more it costs, which caps how much a casual user will attempt. Free Mode appears to move that cost off the individual account and onto platform economics, and the 80% price cut OpenAI reports on GPT-5.6 Luna is what likely makes such an arrangement workable. It also gives OpenAI a consumer-scale showcase for a cheaper model.

Replit is a browser-based development environment whose AI agent writes, runs and deploys applications from natural-language instructions. Free Mode is the new option inside that product, and OpenAI names GPT-5.6 Luna as the model behind it, presenting the mode as a route into software creation for anyone.

OpenAI says it cut the price of GPT-5.6 Luna by 80%, a reduction dated July 30. The same model became available to free ChatGPT users on August 6, per OpenAI's announcements, so both the pricing and the consumer distribution changes landed before the Replit launch.

The announcement was shared exclusively with Fortune, according to press reports, and OpenAI published its own account of the launch on its site. AI coding tools generally meter usage in tokens or credits, and agent runs that rewrite, test and redeploy an application consume more of that budget than a single prompt does.

What Free Mode leaves unspecified

The available material does not set out how long Free Mode runs, whether it carries usage caps, or which Replit account tiers it covers. No figure has been given for how many applications a user can build under it, and no end date for the arrangement has been publicly named.