Core and Pro subscribers get credit-free everyday Agent tasks, with five-hour limits and up to 30 hours of monthly chat on Core.
By Ryan Merket · Published
Primary source: Replit on X
Why it matters #
OpenAI's 80% Luna price cut lets Replit bundle routine agent work into paid plans, shifting AI coding competition from model access toward how much useful work a subscription includes.
Video version #
Replit launched Free Mode, a lower-cost setting for its coding agent powered by OpenAI's GPT-5.6 Luna, giving paid subscribers a way to handle routine prompts and edits without consuming their monthly credits.
Disclosure: RuntimeWire is built with Replit and would not exist without it. We could not be more thankful for the platform.
https://x.com/Replit/status/2090076648276185555 The San Mateo, California-based company announced the model powering Free Mode in an August 19th post on X, after publishing a full product announcement the previous day. Despite the name, Free Mode is available to customers on Replit's paid Core and Pro plans. Core currently costs $25 monthly or $20 per month with annual billing, while Pro costs $100 monthly or $95 per month when billed annually.
The move extends the original bet made by co-founders Amjad Masad and Haya Odeh: removing the technical and economic friction between an idea and working software. Masad began building Replit's precursor while studying computer science and moving between borrowed machines and internet cafes in Jordan. Odeh joined him on the product's design, shaping an interface intended to work for beginners as well as experienced programmers. Masad later worked on JavaScript infrastructure at Facebook and was a founding engineer at Codecademy before building Replit as a company.
A free mode for paid plans
In Free Mode, Replit says chatting, ideation and "everyday tasks" will no longer draw down a subscriber's usage credits. Core and Pro customers can use the setting until reaching limits that reset every five hours, with higher limits for Pro accounts. Core subscribers also receive up to 30 hours of chat per month.
Replit claims the mode lets Core customers create up to 30 times more than before. That figure is a company estimate, and Replit's announcement does not define the workload or baseline used for the comparison. The harder boundary is clearer: Free Mode has usage limits, and Replit Agent can recommend moving a job to a paid mode when the work becomes more complex.
Power Mode, previously called Economy Mode, uses cost-optimized models and consumes credits. Max Mode is aimed at longer builds, larger changes and tasks that require deeper reasoning. Replit also maintains a separate Starter plan, which costs nothing and includes daily Agent credits under a monthly cap. Free Mode therefore functions as an allowance inside paid subscriptions rather than a replacement for Replit's existing free tier.
The naming gives Replit a marketing hook, while the underlying product decision addresses a real constraint in agentic software development. Usage-based billing makes builders consider the cost of every question, planning session and minor edit. That discourages the frequent, low-stakes interactions needed for an agent to become the main interface for a project. Replit is using cheaper inference to absorb those interactions, reserving its credit meter for heavier work.
OpenAI cut the cost of the subsidy
The economics changed on July 30th, when OpenAI cut GPT-5.6 Luna's price by 80%. OpenAI's current API documentation prices GPT-5.6 Luna at $0.20 per million input tokens and $1.20 per million output tokens, with cached input priced at $0.02 per million tokens. OpenAI describes Luna as its GPT-5.6 model for cost-sensitive, high-volume workloads.
Luna is the smallest and least expensive tier in the GPT-5.6 family, below the Terra and flagship Sol models. It supports a 1.05 million-token context window and up to 128,000 output tokens, specifications that make it suitable for carrying substantial project context while handling routine agent work at a lower cost.
OpenAI has already applied the same distribution strategy to its own consumer product. On August 6th, it said Luna would become the default model for free and Go users in ChatGPT, with unlimited text chats. Replit is now using Luna to increase the included consumption on its paid plans, pairing a cheap general model with higher-cost options for work that needs more capability.
For Masad and Odeh, Free Mode pushes Replit closer to the continuous workspace they have spent years building: a place where users can discuss an idea, plan it and turn it into software without moving between a chatbot, editor and deployment service. Whether the mode changes user behavior will depend on the five-hour limits and the point at which Agent pushes a project into Power or Max. The immediate business logic is simpler. OpenAI made intelligence cheaper, and Replit is spending part of that reduction on keeping subscribers inside Replit.