Google AI Studio Free Plan Pricing: Pro, Ultra and API Billing

The Google AI Studio free plan still costs $0 for basic work in the web interface. Google AI Pro and AI Ultra add development access and quota, but neither turns your application's Gemini API calls into an unlimited subscription. Stay on Free until model access or interruptions block your work; budget production API usage separately.

Google AI Studio is Google's browser workspace for testing prompts in Playground and building applications in Build. The important buying question is where the work runs: your development session, or requests made by software you publish.

Verified against Google's live AI Studio documentation and Google AI plans pricing page, November 2026; checked November 4. Prices below are the public US monthly offers. The comparison turns on model access, interruptions during development, and application traffic, without assigning value to unrelated subscription extras.

Google AI Studio Pricing Plans

Free is the starting point; Pro buys more room to develop, and Ultra buys the highest web-interface quota. A paid API project is a separate usage arrangement that can coexist with any of them.

The subscription amounts come from Google's US AI plans page. Its starting Ultra offer includes $40 in monthly Google Cloud credits through the Google Developer Program; Pro includes $10, subject to eligibility and activation. Those credits can offset eligible consumption, but they do not erase the billing boundary.

Treat from as part of Ultra's price. Compare the offer shown for your account and country before subscribing. The public page checked for this article did not supply an annual price for this comparison, so multiplying a monthly price by a year would not establish an annual-plan discount.

Google AI Studio Free Plan: What You Can Build

Free is enough to find out whether your idea works, provided the available models and modest quota fit your experiments. Use Playground to refine a prompt, then Build to turn an idea into an application and iterate on it.

Imagine a founder exploring an appointment-intake tool. Start with invented customer messages, decide what information the tool should extract, and inspect whether the results are useful. A paid subscription adds little while the unresolved question is what the product should do. The sensible stopping point is evidence that access or quota is preventing the next useful experiment.

Which Models Are Free, and Which Are Excluded?

Gemini 4 Argon, Google's frontier model, is explicitly excluded from the Free AI Studio plan. The API pricing catalog separately lists free input and output for Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.1 Flash-Lite. These are examples of available free API options, not a guarantee of unlimited use.

The same catalog marks Gemini 3.1 Pro Preview, Gemini Nano Banana 2.1 image generation and Veo 3.1 video generation as unavailable on the free API tier. Read that column as an API entitlement. It does not tell you whether a paid web subscription includes a particular development experience.

That distinction matters when selecting a model. A founder validating text extraction should first ask whether an available Flash model is adequate. A product whose defining feature requires a paid model should budget for that requirement early instead of designing around the assumption that everything selectable in a browser is free.

Who Never Needs to Pay?

An occasional builder needs no paid plan while their experiments fit the available access and waiting after a limit is acceptable.

For example, a founder sketching internal tools between other projects may get no value from a larger daily allowance. Finish the product definition, revisit a prompt later, and subscribe only when an interruption has a meaningful cost. A successful prototype alone is not a reason to upgrade.

What Pro and Ultra Change Inside AI Studio

Pro adds premium access including Gemini 4 Argon, Gemini Pro, Nano Banana and Lyria, alongside higher quota in Playground and Build's Code Assistant. Ultra provides the highest quota for development. Daily allowances reset rather than replenishing through a rolling window, but Google's plan documentation does not publish a universal requests-per-day count or reset timezone.

Those are the stated benefits in Google's AI Studio subscription documentation. It also excludes Deep Research and Antigravity Preview agents from these plan entitlements: agents inside AI Studio require a paid API key, even for subscribers.

Pro is the reasonable first subscription for someone working in the browser most days. Ultra needs a stronger justification: interruptions that remain expensive after Pro, or a development workload that repeatedly needs its higher access. A plan label cannot answer whether it will keep up with your particular project.

A useful buying test is to record what stopped a normal session. Was it unavailable model access, an exhausted web allowance, or your app's API quota? Only the first two point directly toward a web subscription. Buying Ultra to solve an API-project restriction spends money on the wrong control.

The Pro-to-Ultra Break-Even

Starting Ultra costs $80 more per month than Pro. Its displayed Cloud credit benefit is $30 larger, so even if you consume all that extra eligible credit, the remaining incremental cost is $50. The subscription invoice itself does not fall.

For an illustrative developer who values focused time at $60 an hour, Ultra must save more than 80 minutes of monthly interruption to justify that increment on time alone. With the full additional credit replacing spending that would otherwise occur, the threshold becomes 50 minutes. Neither calculation assumes a published quota multiplier.

On an assumed 20-workday month, Pro's subscription is about $1 per workday and starting Ultra is about $5. That comparison helps value browser access; it says nothing about how many production requests either plan buys. Start with the cheaper option unless your own interruption record supports the difference.

When Does the Gemini API Start Billing You?

A request becomes paid usage when it runs through a paid API project under the applicable model and feature rates. Simply reaching the web allowance does not establish an automatic subscription overage: Google's documented continuation path is to use a Gemini API key with Cloud Billing enabled.

An API key identifies the project receiving the request. Think of the subscription as access to your development workspace and the key as the account number on the application's usage meter. Paying for the workspace does not settle every bill generated by what you build there.

Location does not decide the charge. Google's billing FAQ says that linking a paid API key in AI Studio makes usage through that key billable. A browser tab is therefore not proof that a request is covered by Free or your subscription. A free-tier project can still make eligible free API calls within its own limits.

Build Assistance and App Traffic Are Different Work

The Code Assistant helps create and change an app. The app then makes its own model requests when someone uses a feature powered by Gemini. Google's Build mode documentation explains that web apps use a server-side API key, shared usage counts toward the owner's limits, and paid models can incur charges. A deployed app uses the owner's key for users' Gemini calls; hosting can also cost separately.

For the appointment-intake example, asking the assistant to add a confirmation screen is development work. A visitor submitting a message that your app sends to Gemini is runtime work. If you invite customers, forecast the latter before deciding that the prototype's development cost represents the product's operating cost.

The most useful handoff is a short note attached to the project: which model the app calls, which key pays, and how much a typical interaction consumes. That makes the change from experimenting to serving customers a deliberate budget decision.

What a Small API Workload Costs

For Standard paid API text usage, input/output rates per million tokens are Gemini 3.5 Flash-Lite $0.30/$2.50; Gemini 3.8 Flash $0.75/$3.75 through December 31, 2026, then $1.50/$7.50; Gemini 4 Argon $2/$10 through November 30, 2026, then $4/$20. Output includes thinking tokens. Use Google's API pricing page for the current model catalog and our Gemini pricing guide for the wider pricing context.

A token is a small unit of model input or output. To normalize the options, assume each short text request consumes 2,000 input tokens and 500 billable output tokens, including any thinking. Across 1,000 requests, that is 2 million input tokens and half a million output tokens.

At the currently listed rates, that workload costs $1.85 on Flash-Lite, about $3.38 on Flash, or $9 on Argon. These are calculated examples, not measured application bills. They exclude tool calls, caching charges, hosting and taxes; real output length, reasoning and retries can change consumption. The scheduled rate changes make the same Flash workload $6.75 and Argon workload $18.

The comparison is useful because the alternatives solve different problems. For occasional prompts, use Free. For daily browser development, value the subscription against avoided interruptions. For a small app, estimate requests and pay the API project directly. A subscriber building a customer-facing product may rationally need both.

Cloud Credits, Prepayment and Other Costs

Eligible subscription credits reduce an API bill only when they are attached and usable. Google's Developer Program billing FAQ says prepaid accounts need an active paid balance before eligible Cloud credits can be consumed; those credits are used first, and stop being consumed when the prepaid balance reaches $0.

This matters for the economics. If Pro's entire $10 benefit replaces API usage you would otherwise pay for, its additional cost over API-only spending is $9.99. For starting Ultra with the full $40 benefit used, it is $59.99. If you do not consume the credit, its unused face value has not saved you cash.

Google One AI credits are a different system and are not supported in AI Studio. Also keep the subscription benefit separate from the $300 Google Cloud free trial: the current Gemini billing documentation excludes that trial from Gemini API usage. An account showing a promotional balance is not enough; check which service and billing account can spend it.

Credit Expiry and Refunds

Purchased API Prepay credits expire after 12 months and are generally nonrefundable; an eligible switch to Postpay is a documented exception. Buy a balance that matches expected use instead of treating it as money that will remain available forever.

The Developer Program benefits FAQ separately describes Cloud credits expiring one year after grant and explains how to apply recurring benefits to a chosen billing account. Check the grant attached to your account. Purchased API credit and a subscription benefit are different balances with different origins.

Canceling the subscription is also different from recovering its cost. The plans page says you can cancel anytime, while Google One's published refund policy generally makes purchases nonrefundable in most regions, with exceptions depending on location and purchase channel. Do not interpret cancellation as a free trial or an automatic prorated refund.

Free Usage Has a Data Trade-Off

For a US builder, Google's unpaid-service terms allow prompts and responses to be used for product improvement and human review. The Gemini API data-use terms specify different treatment for the EEA, Switzerland and UK. Start experiments with synthetic examples and check the applicable terms before using customer information.

That changes the upgrade decision for the intake tool. A demo using invented appointment requests and a live workflow containing personal customer messages are different operating situations, even if they produce the same token count.

Set Up Paid API Usage Deliberately

Choose the project that should pay before connecting customers or running unattended jobs. This keeps the next charge attributable to a workload instead of a forgotten experiment.

1. 1

   Identify the project's billing state

   In AI Studio's API keys or Projects page, locate the relevant key and project. Check its Billing Tier before assuming a subscription covers its calls.

2. 2

   Enable the appropriate payment route

   Use Set up billing for the project. Prepay setup requires at least $5; eligible accounts may use Postpay. Check the assigned billing plan and activate any eligible developer credits on the intended account.

3. 3

   Set a project limit and inspect consumption

   On the Spend page, use Monthly spend cap and Edit spend cap. Project caps are experimental: billing delays and long-running work can cause overages. Review usage after a representative interaction before opening access widely.

These controls and payment conditions are documented in Google's Gemini API billing guide. Do not budget to the exact edge of a balance: a depleted prepaid account stops paid service rather than automatically switching the app to Free.

Questions About AI Studio's Free Plan

Google AI Studio limits free tier

There is no single request count that safely describes every free workload. Separate the web plan's daily allowance from a project's model-specific API rate limits. Inspect the active model and account limits; an old requests-per-minute figure is not a universal entitlement.

Google AI Studio free credits

Free access, Google One AI credits, subscription Cloud credits and purchased API credits are different things. For API budgeting, count only a benefit confirmed eligible and activated for the billing account that pays for your app.

Google AI Studio Tier 1

Tier 1 is a Gemini API usage tier reached through billing setup. It is not Google AI Pro. An API tier governs the project's API usage arrangement; a subscription changes supported work in the AI Studio web interface.

Google AI Studio tiers

Use Free, AI Pro and AI Ultra when comparing web plans. Use the API project's separate tier when diagnosing API limits or charges. Before upgrading, identify which of those two systems produced the restriction.

Google AI Pro Student Offer

Google's plans page currently advertises AI Pro free for one year for eligible students, with availability and terms varying by country. Verify the offer presented to your account. A student subscription should still be evaluated against the separate API billing rules before using it as the budget for a public application.

Stay Free, Subscribe, or Use a Paid API Key?

Stay on Free for occasional Playground experiments and early Build prototypes when you can use the available models and tolerate pauses. The buying trigger is a specific blocked task, not the existence of a paid tier.

Subscribe to Pro when browser development is regular enough that access restrictions interrupt valuable work. Move to Ultra only when the additional development allowance earns its higher cost in your own schedule. Treat Cloud credits as a discount on usage you genuinely expect, not a reason to invent consumption.

Use a paid API key for production-scale app traffic, paid-model API requests or agents such as Deep Research inside AI Studio. A small eligible API experiment can remain free, but a production plan needs an explicit usage budget. Keep a subscription alongside the API only if its development benefits also justify the spend.

For your next work session, inspect the key attached to the app, estimate a representative interaction, and note the first limit you actually hit. Then buy the smallest change that resolves it. That is a better starting budget than choosing Ultra before you know whether your constraint is development access or customer traffic.

Use the AI business workflow audit checklist to identify the workflow worth building before committing to a larger plan.