Gemini API Free Tier 2026: Limits, Free Models, and Free API Keys

Yes, the Gemini API has a free tier, and creating an API key costs nothing. Free inference requires both an eligible project and a model and feature offered on the Free Tier. Google's pricing page lists free input and output for models such as Gemini 3.8 Flash, 3.6 Flash, Flash-Lite variants and Gemini 2.5 Pro. But Google's troubleshooting guide explicitly says not all projects or accounts qualify. A free key and a free model are not enough to establish access.

Once your project is eligible, read its active limits in AI Studio. Limits apply per project, not per key. Grounding also needs a separate check: Gemini 2.5 Flash and Flash-Lite have free Search and Maps allowances in their Standard pricing tables, while other models and modes differ. The details below were checked on October 27, 2026; they are documentation checks, not a test of your account's quota.

Quick answer

The real change is where the limit lives

Older Gemini free-tier posts, videos and examples often show a static RPM/RPD table. Google's current rate-limits documentation directs developers to AI Studio for active per-model limits. A published figure from an earlier date is not a promise about your project today.

The mechanics still matter. Limits are usually measured across requests per minute (RPM), input tokens per minute (TPM), and requests per day (RPD). Exceeding any one can trigger an error while the others still have room. Limits apply per project, not per API key, and RPD resets at midnight Pacific time. Some models have additional dimensions, such as images per minute or tokens per day.

There are three separate checks:

1. Model and feature: the pricing page says whether the exact model, mode and tool have free pricing.
2. Project access: project status and eligibility guidance explain whether your account/project can use that free route.
3. Capacity: AI Studio's rate-limit view shows the selected project's active limits.

For example, a free input/output row cannot override a project's “Free tier unavailable” status. A project that is eligible can still exhaust its daily quota. The first needs an access decision; the second needs a quota or traffic decision.

Which Gemini API models are free-capable now?

The following are selected entries from Google's pricing page, checked October 27, 2026. “Free of charge” describes the listed input/output cells. It does not guarantee project eligibility, every tool, or every service mode.

The 2.5 Flash grounding exception

On Standard, the Gemini 2.5 Flash and 2.5 Flash-Lite tables list:

• Google Search grounding: free up to 500 requests per day, shared between these two models.
• Google Maps grounding: 500 RPD in the Free Tier column for each model.

This does not give every Gemini model 500 free requests per day, or add 500 requests to whatever quota AI Studio shows. The allowance belongs to the specified grounding feature; the project must still be eligible and stay within all applicable limits. The 2.5 Pro tables and the Gemini 3.8 Flash Standard table, for example, mark free Search/Maps grounding unavailable. Other routes, including Live, have their own entries.

Paid-tier included grounding usage is also a different offer. The pricing page lists 5,000 free Search requests per month shared across Gemini 3 and newer models on the paid route, followed by $14 per 1,000 Search requests. Model inference charges still apply, and a Gemini request may trigger more than one Search request. That monthly inclusion does not turn a paid project into a Free Tier project.

For a sense of paid inference cost, Gemini 3.8 Flash Standard is $0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026, then $1.50 and $7.50 from January 1, 2027. Gemini 3.1 Pro Preview starts at $2.00 input and $12.00 output per 1M tokens for prompts up to 200,000 tokens. These are the listed USD rates, before any separately priced tools.

If your task is paid image generation, the Nano Banana API pricing guide compares model and output-size costs. After setting up paid billing, the Nano Banana API tutorial covers model IDs and first-call code.

Is a free Gemini API key unlimited? Auth keys and project limits

The key is free, but it is not unlimited. Google's API key documentation says every key belongs to a Google Cloud project. New AI Studio users get a default project and key after accepting the terms; existing Cloud users can import their projects. Neither flow guarantees that the project qualifies for free inference.

Since May 28, 2026, new keys created in AI Studio are authorization (auth) keys, bound to a service account and restricted to the Gemini API by default. Unrestricted standard keys are rejected; standard keys with appropriate explicit restrictions can still work. Keep credentials on the server and follow Google's key creation and security instructions when importing or replacing a key.

An offer advertised as a “free unlimited Gemini API” should not be confused with Google's official allowance. If requests go through a third-party endpoint, check whose account pays, what limits the provider imposes and how it handles prompts. A working request through that endpoint says nothing about your Google project's free-tier eligibility. Student status likewise does not replace the project, model and quota checks.

Why a free model may still be unavailable to your project

Google says explicitly that not all Google Cloud projects or accounts are eligible for the Gemini API free tier. If an error says “Your project has been denied access,” or AI Studio shows “Restricted,” “Free tier unavailable,” “Payment needed” or “Unavailable,” use the project restriction guidance to distinguish these cases:

The same documentation warns that models.list only checks whether the key exists; it does not check the project's tier eligibility or billing status. A 200 OK model list therefore does not prove that generateContent, streamGenerateContent or interactions will work. When validating your own integration, distinguish a successful metadata request from a successful inference response on the intended project, model and billing route.

Why new API keys do not add quota

Google AI Studio can show many keys, and a Google Cloud project can hold more than one key. That does not mean every key brings its own free allowance.

The API key is an access credential. The quota bucket belongs to the project. If staging, internal demos, a cron job, and a developer laptop all use different keys inside the same project, they still draw from the same project-level rate limit. When a request returns 429, creating another key in that project hides the problem rather than fixing it.

Use separate projects only when the separation is real:

• different billing ownership,
• different environments with separately managed quota,
• a genuine security or permission boundary,
• a team structure that can monitor usage separately,
• and a policy or compliance reason for the split.

Do not split projects merely to get around a free-tier cap. It complicates monitoring and does not resolve an eligibility restriction on the intended project. If the workload needs more capacity, the real options are a different model, a paid tier, caching, request shaping, or an official quota increase.

How to check the live limit before you build around it

Before a serious prototype, check the exact project your application uses:

1. Open the pricing page and confirm the model, service mode and requested tools have the Free Tier entries you need.
2. Open AI Studio Projects and check the project's tier, status and linked billing account. Resolve any eligibility or billing status before estimating throughput.
3. Open the rate-limit page for that project and model. Note RPM, input TPM, RPD and any additional dimensions.
4. Record the model, project, tier, timestamp and observed limits before a demo, launch or load test.
5. Recheck after changing model, project, billing account or traffic shape.

If AI Studio shows too little RPD for a demo day, decide whether to reduce calls, use another suitable model or pay for capacity. If your project is ineligible or the feature has no free entry, a retry library cannot create that entitlement. Google's rate-limit documentation also says stated limits do not guarantee actual capacity.

What a 429 means on the Gemini API free tier

A 429 signals a limit problem, but the full error identifies which one. It is not proof that Google removed the Free Tier, and it is not automatically a billing problem. Google's error guidance and rate-limit documentation distinguish request, token, daily and applicable spend limits.

If quota details show limit: 0, there is no usable allowance for the named quota in that request's context. The zero alone does not explain why. Match the quota metric and model to the intended project, inspect AI Studio's status, and check the model/feature pricing. A paid-only image model called through the free route has no free image allowance; a model with free pricing still needs an eligible project. Do not label every zero limit as a paid-only model or every 403 as the same restriction.

If the error and AI Studio figures still disagree, collect the project identifier, exact model, operation, timestamp, quota metric, displayed limit and recent usage for support. Remove the API key and sensitive prompts from anything you share. Do not repeatedly create projects or keys as a substitute for finding the cause.

When free tier is enough

The free tier works when the workload is closer to evaluation than infrastructure. Good fits include:

• prompt exploration,
• classroom or tutorial code,
• low-risk prototypes,
• internal demos,
• model comparison,
• small personal tools,
• short-lived proof-of-concept agents,
• and early cost and performance experiments.

Even then, shape your traffic. A prototype that queues bursts, caps output tokens, caches repeated work, and logs which limit caused each 429 is much easier to move to paid capacity later. A prototype that assumes "free means unlimited until launch week" has the wrong architecture.

The free tier becomes a poor fit when the product needs capacity beyond the available quota, a paid-only model or feature, or a service/data-use arrangement the unpaid route does not provide. Paying can increase available capacity, but it does not remove rate limits or guarantee uptime.

Data use: the European exception is separate from app availability

Under the general Unpaid Services rule, Google uses submitted content and generated responses to improve its products, and human reviewers may read and annotate them. The terms tell users not to submit sensitive, confidential or personal information under that rule. Paid Services do not use prompts and responses for product improvement, but Google still logs them for a limited period for safety, security and required legal purposes. Paid handling is not zero retention. Source: Gemini API terms.

Two European provisions apply to different subjects:

Both provisions appear in the Gemini API terms. Google's billing FAQ also says free and paid tiers are available in many regions including the EEA, UK and Switzerland. That is why “Europe has no free API access” and “all free-tier data is used for improvement” are both too broad.

When to move from the free tier to Tier 1: billing, caps, and spend limits

Move to a paid tier when you need a paid-only model or feature, more capacity than your project has, the paid route for an otherwise ineligible project, or Paid Services for your application's users or data requirements. Linking active billing changes API usage to the paid route; it does not unlock a free allowance beneath paid usage.

In AI Studio, choose Set up billing for the intended project and complete the flow. If assigned to Prepay, add the required credits; Google's setup guide specifies a minimum of $5 or the equivalent in other currencies. Some accounts use Postpay or are offered a choice, so follow the account's actual billing state. Source: billing setup.

Sources: rate limits and billing, checked October 27, 2026. The spend-window limits depend on billing history and account standing. Paid projects inherit their billing account's tier and share its monthly cap; ordinary rate limits still apply per project. The criteria do not guarantee approval or a particular per-model quota. Check the resulting tier and limits in AI Studio after billing changes.

Billing and prepay can be the hidden limiter

On Prepay, purchased credits expire after 12 months. If the balance is depleted, Gemini API service stops across projects funded by that billing account; the projects do not automatically fall back to Free. Add credits to resume paid service, allowing for payment and balance-update delays. An overdue balance elsewhere on the linked Cloud Billing account can also interrupt access, even when Gemini prepaid credits remain. Source: Google's billing guide.

If production depends on paid capacity, assign an owner for top-ups and monitor the balance, account status and spending limits. Auto-reload is an option, with its own charging settings to review. For the paid qualification path beyond Tier 1, the Gemini API Tier 3 guide explains the spend and account-age requirements.

Gemini API free tier regions: where you can use it

Check Google's available-regions list before building on the Developer API or AI Studio. As checked October 27, 2026, it includes the United States, Japan, South Korea, Spain, Taiwan, the UK and Switzerland, among others. Mainland China, Hong Kong and Russia are not listed. Linking billing does not override that service's regional availability.

Being in a listed region also does not prove that a particular account or project qualifies for the Free Tier. Apply the project eligibility check separately, and use the two European rules above when deciding how data is treated and whether your application must use Paid Services. For hosted environments, deployment location can matter: Google's regions page specifically says Colab restrictions use the instance's region.

Where a fallback provider fits

Check Google's own routes first. A multi-provider gateway may be useful when your application needs several model families behind one integration or a separately evaluated fallback. It has its own billing, model support and data-handling terms; it does not restore your Google project's free quota.

For teams that want one gateway across providers, laozhang.ai is one option to evaluate once the official Gemini API contract is clear. Compare any provider on the same terms as Google direct: price per successful request, latency, concurrency behavior, retry logs, and data handling.

Disclosure: laozhang.ai is our own service. Its prices are platform pricing, not an independent comparison.

That order matters:

1. Confirm the exact model, mode and feature pricing.
2. Confirm project eligibility and the active quota in AI Studio.
3. Shape traffic and reduce avoidable 429s.
4. Decide whether paid access, a quota increase or a separately evaluated provider fits the workload.

Launch checklist

Before you depend on the Gemini API free tier for anything shown to other users, check that:

• The exact model, mode and tools have the Free Tier entries you need.
• The intended project is eligible and shows enough active quota for the workload.
• The key has the right restrictions and permissions and belongs to that project.
• The project's tier and billing state are known, including any required Prepay balance.
• Request code has queueing, bounded retries, token caps and useful error logging.
• Monitoring distinguishes RPM, TPM, RPD, eligibility and billing failures.
• The applicable data-use terms fit the material you send, including any regional exception.
• API clients offered to users in the EEA, Switzerland or UK use Paid Services.
• A capacity and billing decision exists before real users depend on the prototype.

FAQ

Can I get a Gemini API key for free?

Yes. Google AI Studio creates the key at no cost, and the key documentation describes default project creation for new users. A key is a credential. Free inference still requires an eligible project and free-capable model, mode and feature; it is not guaranteed merely because key creation succeeds.

Is the Gemini 2.5 API free?

The pricing page lists free Standard input/output for Gemini 2.5 Pro, Flash and Flash-Lite. The 2.5 Flash Native Audio and Flash Preview TTS entries also have free input/output, while 2.5 Pro Preview TTS and Flash Image do not. Check project eligibility and current model status as well as pricing.

Is Google Search grounding free with Gemini?

For gemini-2.5-flash and gemini-2.5-flash-lite on Standard, Google's pricing tables list up to 500 free Search requests per day shared between those models. Their Maps rows list 500 RPD. These tool allowances do not establish every model's free request quota; project eligibility and other limits still apply.

Which Gemini models are free in the API right now?

As checked October 27, 2026, Google's pricing page includes free Standard input/output for Gemini 3.8 Flash, 3.6 Flash, 3 Flash Preview, 3.5/3.1/2.5 Flash-Lite, 2.5 Flash and 2.5 Pro. Selected Live, TTS, embedding, robotics and Gemma entries also have free pricing. Use the exact model's table rather than assuming every model in a family or every feature is included, and check that your project qualifies.

Is Gemini 3.1 Pro free through the API?

No. gemini-3.1-pro-preview is not available on the Free Tier. Seeing the model in AI Studio or a model-list response does not grant free backend inference.

Does each Gemini API key get its own free rate limit?

No. Google's rate-limits documentation says limits apply per project, not per key. Creating another key in that project also does not change its Free Tier eligibility.

Where do I see my actual Gemini API free-tier rate limit?

Open AI Studio's rate-limit page with the correct project selected. Check the model's RPM, input TPM, RPD and any additional dimensions. A number shown for another project or in an old tutorial is not your allowance.

How much is 1 million tokens in Gemini?

On an eligible Free Tier project, the model's free input/output rows cost nothing within applicable limits. On paid Standard, Gemini 3.8 Flash pricing is $0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026, rising to $1.50 and $7.50 from January 1, 2027. Input and output are billed separately, and tools may add charges.

Why does my free-tier 429 say “limit: 0”?

It means the named quota provides no usable allowance in that request's context; the number alone does not identify the cause. Compare the quota metric, model, mode, actual key/project and AI Studio status. A paid-only model has no free allowance, but a free-capable model still requires an eligible project. Use the official troubleshooting guide before assuming another key or more retries will fix it.

Why can I list models but not generate a response?

Google says models.list validates the key's existence without checking project tier eligibility or billing. Its 200 OK response therefore does not guarantee inference access. Check the failing operation's error and the project's eligibility and billing status. Source: project restriction troubleshooting.

Can I use the free tier in Europe?

Free access exists in supported European regions, subject to account/project eligibility. For users located in the EEA, Switzerland or UK, Paid Services data-use terms apply even to unpaid quota. Separately, developers must use Paid Services when making API clients available to users in those regions. The terms and billing FAQ describe these distinct conditions.

Can I use the Gemini API free tier for production?

It can fit low-risk prototypes and evaluation when eligibility, quota and the applicable terms permit the use. For an application serving real users, choose paid access when required by its region, model, feature or data-use needs, and plan around actual capacity. Neither a free key nor a paid tier removes limits or guarantees uptime. Start with the service terms and your project's active limits.