Replit turned on Free Mode for its coding agent on Tuesday. The new tier runs everyday tasks on OpenAI’s GPT-5.6 Luna model without touching the usage credits that meter every other mode on the platform. Replit’s Michele Catasta signed the announcement. He frames the launch as an answer to a familiar complaint: capability is now cheap, but tracking which model to use and how many credits remain is not.
The stakes sit in who pays for the compute. Under Free Mode, Replit itself absorbs the inference cost of every GPT-5.6 Luna call that falls inside its everyday-task definition. That expense gets folded into the existing $20 monthly Core subscription instead of metered per request. The arrangement only works if OpenAI’s per-token price for this model has fallen far enough for Replit to absorb it at scale. It is a bet on infrastructure economics, not a gift to users.
Replit says Core subscribers can now create 30 times more inside Free Mode, plus up to 30 hours of chat per month. The company did not publish how it calculated that multiplier or what baseline it used. Pro subscribers get higher usage limits than Core, though Replit did not specify by how much.
Free Mode is not unlimited. Usage limits reset every five hours, and once a task grows complex enough, Replit’s agent routes it into one of two paid modes instead. Power Mode, the tier Replit renamed from Economy Mode, runs lower-cost models for routine work. Max Mode calls on stronger, higher-performance models for the heaviest jobs: large changes, deep reasoning, long builds. Both non-Free modes still draw down credits the way the whole platform did before Tuesday.
That renaming is telling. Power Mode did not arrive as new infrastructure: it is Economy Mode with a new label, sitting behind a new free lane. Replit’s actual product change is narrower than the announcement suggests. The company built a router that decides, task by task, whether a prompt is cheap enough to give away or expensive enough to bill. It wrapped a redesigned interface around that router, so users move between chat, tasks, and full builds without switching tools.
The redesigned interface is the second half of the launch. Replit describes it as a single surface that carries a user’s context from a quick question to a shipped build. That positions the product less as a coding editor and more as a standing collaborator across sessions. Replit disclosed no adoption or retention numbers from that early period.
For teams paying for Replit Core today, the open question is not whether Free Mode is generous. It is whether their actual workload fits inside Replit’s definition of everyday, or whether it routes into Power or Max Mode by the second week. If it does, the bill looks identical to what they paid before. Anyone evaluating a 2026 Replit contract should track how often their team’s real usage trips the five-hour reset before assuming Free Mode changes the unit economics of their build.
Replit announced Free Mode in a company blog post published August 18, 2026.
