LLamjinLab
AI NewsWords 1380Read time≈ 4 min

Replit Launches Free Mode With GPT-5.6 Luna and Plan-Based Usage Limits

Replit Free Mode uses GPT-5.6 Luna for credit-free Agent work, with plan-based limits and paid modes for more complex tasks.

AIAgentOpenAILLM

Replit has launched Free Mode, a new default path for using its AI Agent without consuming Power or Max Mode credits while a user’s plan-based allowance remains available. The mode is powered primarily by OpenAI’s GPT-5.6 Luna and is intended for everyday chat, planning, initial builds, and smaller changes.

The product change reorganizes Replit Agent around three modes: Free, Power, and Max. It also changes where metered usage begins. Routine work can remain inside Free Mode, while users move to credit-consuming modes when they need a manually selected model, more capability, or continued access after reaching their Free Mode limit.

Replit says Core subscribers can create up to 30 times more than before and receive as much as 30 hours of chat per month through Free Mode. That multiplier is Replit’s own comparison; the company has not published the workload, prompt mix, or measurement methodology behind it.

1. What Free Mode Includes

Free Mode can chat, create, and modify projects without drawing from the credits allocated to Power or Max Mode. Replit’s documentation recommends it for exploring an idea, starting a project, producing a first pass at a feature, planning work, and handling smaller changes.

It is not unlimited. Allowances depend on the user’s plan, and Replit shows the remaining amount and refresh time under Settings → Usage. The launch announcement says usage limits reset every five hours and that Pro customers receive larger limits than Core customers.

Free Mode always uses Replit’s Auto routing system. Users cannot choose a specific model while the mode is active. Manual model selection is reserved for eligible Core and Pro users working in Power or Max Mode.

The mode also excludes Replit Routines, which automate recurring tasks. Routines must run in Power or Max Mode and require a budget for each run. This preserves metered billing for unattended or repeated execution even when ordinary interactive work can be performed within the Free allowance.

Replit says it warns users before a request shifts into a mode that incurs usage costs. A user who reaches the Free Mode allowance can wait for the displayed reset or, on an eligible paid plan, continue in Power or Max using monthly credits.

2. GPT-5.6 Luna Provides the Economic Foundation

OpenAI identifies GPT-5.6 Luna as the primary model behind Free Mode. Luna is the smallest and least expensive tier in the GPT-5.6 family, below the balanced Terra model and flagship Sol model.

OpenAI cut Luna’s API price by 80% on July 30, 2026. Its current API price is $0.20 per million input tokens and $1.20 per million output tokens, down from its initial GPT-5.6 pricing of $1 and $6 respectively. OpenAI attributed the reduction to efficiency improvements rather than an increase in available computing capacity.

That price change is central to Replit’s launch. Replit CEO Amjad Masad said the lower model cost made it practical to offer Free Mode to millions of users. Replit President and Head of AI Michele Catasta separately told Fortune that many routine tasks do not require a frontier model and can instead be handled by a faster, less expensive system.

Free Mode is not necessarily a Luna-only execution path from start to finish. OpenAI says Replit can route a task to GPT-5.6 Sol when more advanced reasoning is required, then return the interaction to Luna while preserving project context. Fortune reported that such rerouting lasts only as long as needed to complete the more demanding portion of a task.

This routing model matters because it separates the interface from the underlying model choice. Users select Free Mode, but Replit decides when a request can remain on Luna and when a stronger model is required. Developers who need deterministic control over the selected model must use Power or Max Mode instead.

3. Replit Has Reorganized Its Agent Modes

Before this release, Replit’s mode structure included Lite, Economy, and Power, with additional high-performance options changing over time. Economy was the cost-optimized choice, while Power targeted complex work and larger codebases.

The new structure changes both the names and their positions:

  • Free Mode is the allowance-based option for chat, planning, early builds, and everyday work.
  • Power Mode replaces Economy Mode at the same cost and is now the credit-consuming, cost-optimized choice.
  • Max Mode is the higher-capability option for complex features, larger changes, deeper reasoning, and production-grade work.

Replit says Power provides the same or better performance than the former Economy Mode without changing its price. Core and Pro subscribers can disable Auto in Power or Max and select from the models available to their plan.

The company has also redesigned the surrounding interface to reduce the boundary between conversation and implementation. Replit says Agent can retain project context while users move from questions and analysis into building, rather than requiring them to copy information into a separate coding tool.

For a typical workflow, that means a user can explore requirements in Free Mode, start an implementation, and move to Power or Max only when the task requires a stronger model or exceeds the available allowance. The practical change is therefore not simply cheaper inference: Replit is placing model routing and cost escalation inside one continuous project session.

4. “Free” Refers to Usage Within an Allowance

Replit’s naming can be confusing because Free Mode is different from the company’s free Starter subscription.

The current pricing page lists Starter at no charge, with free daily Agent usage. Core is advertised at $20 per month when billed annually and includes more free Agent usage plus $25 for the most powerful models. Pro is shown at $95 per month when billed annually, with still more free usage and $100 for higher-capability models.

Replit’s current documentation indicates that Starter users can use Free Mode but must upgrade before accessing Power or Max. Core and Pro users receive larger Free Mode allowances and can continue through their monthly credits after the allowance is exhausted.

This is more expansive than Fortune’s initial launch report, which described Free Mode as requiring a $20 Core or $100 Pro subscription. Replit’s live documentation and pricing page now show Free Mode as the entry path across plans, with the quantity of included usage and access to paid modes varying by subscription.

The distinction is important for evaluating the “30X more” claim. Replit is not making all Agent work unmetered, nor is it removing credit-based pricing. It is creating a larger allowance for tasks that can be served through its automated routing system, while retaining credits for model selection, higher-complexity work, routines, and continued use beyond the Free Mode limit.

For existing customers, the immediate effect is a lower marginal cost for iterative work such as refining requirements, asking project-specific questions, testing an idea, or making an initial change. Teams still need to monitor when work moves into Power or Max, particularly for long builds and automated tasks.

Frequently Asked Questions

Is Replit Free Mode completely unlimited?

No. Its allowance depends on the user’s plan. Replit displays the remaining allowance and reset timing under Settings → Usage.

Which model powers Free Mode?

OpenAI’s GPT-5.6 Luna is the primary model. Replit can route a more demanding step to GPT-5.6 Sol and then return to Luna while preserving context.

Does Free Mode consume Replit credits?

It does not consume Power or Max Mode credits while the Free Mode allowance is available. Credit-consuming modes remain available to eligible paid subscribers.

Can users select a model in Free Mode?

No. Free Mode always uses Replit’s Auto routing. Manual model selection is available to eligible Core and Pro users in Power or Max Mode.

Can Replit Routines run in Free Mode?

No. Routines run in Power or Max Mode and require a per-run budget.

Sources

Share

Share this article