Replit, known for letting people build applications without writing code, has introduced a new way to use its Agent called Free Mode. It runs on OpenAI's GPT-5.6 Luna. Chatting, asking for feedback, and running light analysis no longer consume usage credits, which significantly expands how much a monthly subscription can produce. The announcement came on August 19, 2026.

No more watching the credit meter

When people build with AI, the time spent managing the tooling often exceeds the time spent actually building. Which model to choose, how to pass context, and how much usage is left. Replit went after that third problem.

In Free Mode, answers to questions, feedback on ideas, suggestions, and project analysis do not draw down credits. Because Agent holds the full context of a project, users can work through an idea in the same window and then move straight into implementation. The habit of brainstorming in a separate AI tool and pasting the results back disappears.

Replit says the change lets Core subscribers create 30 times more than before, with up to 30 hours of chat per month. Core costs 20 USD per month (about 3,200 yen).

1 USD = 159 JPY (as of August 24, 2026)

Heavier work moves to other modes, and Replit tells you first

Not everything stays inside Free Mode. Usage limits apply, and they reset every 5 hours. When a limit is reached, or when the task outgrows what Free Mode is meant for, work shifts to another mode. Pro subscribers get higher limits than Core.

There are two of them. Power Mode, previously called Economy Mode, uses cost-optimized models and balances speed against quality. Replit says it delivers the same or better performance than the old Economy Mode at the same cost. Max Mode uses higher-performance models and suits large changes, deeper reasoning, and longer builds.

For requests that need more advanced reasoning, Replit routes the work to GPT-5.6 Sol and then returns to GPT-5.6 Luna once it is done. Project context is preserved across that hop, so the conversation does not restart. Agent surfaces a prompt before switching, which keeps users from unknowingly running on a more expensive mode.

The interface was rebuilt too

The announcement also covers a redesigned interface. The goal is to connect quick chats and small tasks with complex builds in a single place, so questions, data analysis, design, and implementation sit on a continuum. The intent is clear enough: Replit wants to be where the working day starts.

Price cuts made the allowance possible

Free Mode exists because of economics rather than a technical breakthrough. OpenAI frames the price performance of the GPT-5.6 series, along with its recent price cuts, as what made an offer of this scale viable. Amjad Masad, Replit's co-founder and chief executive, has said the cuts are what allowed the company to put Free Mode in front of millions of users.

As per-token costs fall, uses that never penciled out become practical options. Free Mode is a clear illustration of how that shift reaches end users. Replit has built on OpenAI models since the GPT-3 era, and both companies have long stated the same goal of making software creation available to anyone regardless of technical background.

One caveat: despite the name, Free Mode is not free for everyone. It requires a paid Core or Pro subscription, and within that plan it is the mode that does not spend additional credits.

Summary

Replit's Free Mode uses GPT-5.6 Luna to move everyday interactions outside of credit consumption. Core subscribers get 30 hours of chat per month, and Replit describes the output ceiling as 30 times higher than before. Heavier work shifts to Power Mode or Max Mode, with routing to GPT-5.6 Sol where needed. A model price cut has been converted directly into how much users can make.