Gemini API for free: what the free tier gives and its limits

The query "Gemini API key free" usually means one of two things: someone wants to test an idea without spending money, or someone hopes to avoid paying for years. Google has a free tier for the first case, and it does not fit the second. Below we go through what the free tier includes according to Google's documentation, where its boundaries are and when paying per token becomes the better option.

We read the ai.google.dev pages (prices, limits, regions) on October 18, 2026. Google changes terms from time to time, so check the current pages before deciding.

What the Gemini API free tier is

Google's pricing page lists three levels for the Gemini API: Free, Paid and Enterprise. The free level gives you:

• input and output tokens that are not charged;
• access to Google AI Studio;
• limited access to some models;
• content that may be used to improve Google's products.

The last point matters for projects that handle other people's data. In the gemini-3.8-flash price tables, the "Used to improve our products" row says "Yes" for the free tier and "No" for the paid tier. That means the free tier is a poor fit for client data, contracts, correspondence and any personal information: read Google's terms and your company's policy first.

Another limit: not every model is free. On the pricing page some models show "Not available" in the Free Tier column. Google changes which models are free, so look at the column for the model you plan to use.

Limits: RPM, TPM and RPD

The free tier is limited by speed and volume rather than money. The rate limits documentation lists three dimensions:

• RPM, requests per minute;
• TPM, input tokens per minute;
• RPD, requests per day.

Exceeding any of them produces an error even if the others are fine. If your RPM limit is 20, the 21st request within a minute is rejected even though the daily limit is far away.

What to remember from the documentation:

• limits apply to a Google Cloud project, not to an individual key, so a second key in the same project gains nothing;
• the daily RPD counter resets at midnight Pacific time;
• limits depend on the model, and experimental and preview models are stricter;
• the documentation text does not publish concrete numbers, AI Studio shows them for your project; the limits are also marked as not guaranteed.

So the honest answer to "how many requests per day are free" is: look at your project in AI Studio and do not build those numbers into production. We return to this in the section on when to switch to paid.

Regions and age

The free tier works only where the Gemini API and Google AI Studio are available. Google's Available regions page lists countries and territories, and at the time of our check Russia is not in that list (the page was last updated in April 2026). Besides region, access is restricted by age (18+) and age verification on the Google account.

This means that from Russia, getting and using a Google key directly often does not work, and the cause is not limits but access to the service itself. We will not recommend schemes for getting around that restriction. The legitimate route from Russia is a gateway: Gemini is reached through Zukko, with payment by SBP, MasterCard or crypto. More on that below.

Free for prototyping: what is realistic

The free tier makes sense when all of the following hold:

1. You are in a supported region and can create a key in AI Studio.
2. The request data is not confidential and can be given to Google to improve its products.
3. The load is occasional: a notebook with experiments, a study project, a hypothesis check.
4. You are fine with requests starting to return 429 RESOURCE_EXHAUSTED when you hit a limit.

For everything else the free tier stops being convenient. Typical signs: users of your bot get errors at peak hour and nothing works at night until the daily limit resets; you cannot guarantee a response time to a customer; you need to process data that must not be shared for product improvement.

When paying per token is simpler

At Google the paid tier requires linking billing and topping up the account; according to the billing page in Google's documentation, the minimum prepayment when moving to the paid tier is $5. Tiers 1 to 3 open as spending grows and give higher limits. That works for companies with a payment history in Google Cloud, but not for everyone.

An alternative for a developer is paying per token through a gateway. In Zukko, gemini-3.8-flash costs $0.125 per 1M input tokens and $0.6 per 1M output tokens as of October 18, 2026. Top-ups start at $1 and there is no subscription.

Let us price a one-off job: classifying 10,000 short tickets, 400 input and 50 output tokens each.

• input: 10,000 × 400 = 4M tokens, 4 × 0.125 = $0.50;
• output: 10,000 × 50 = 0.5M tokens, 0.5 × 0.6 = $0.30;
• total: 0.50 + 0.30 = $0.80.

At the official prices from our table (input $0.75, output $3.75) the same job would cost 4 × 0.75 + 0.5 × 3.75 = $4.875. This calculation ignores caching and reasoning tokens: Gemini 3.x has reasoning on by default, and per Google's documentation it is billed as output tokens. On harder prompts the total is higher, so run a hundred of your own requests and check the spend in the dashboard ("Usage").

For comparison, a monthly bot budget: 100 requests a day with 3000 input and 700 output tokens each. Per day that is 0.3M × 0.125 + 0.07M × 0.6 = 0.0375 + 0.042 = $0.0795, and over a month (30 days) about $2.39.

A word on "free": Zukko has no free tier. The only addition to a balance is a one-off $0.50 for subscribing to the @zukkopro Telegram channel, once per Telegram account. It is not free access, just a small reserve for testing the connection. You do not need the subscription: you can top up from $1.

How to move to pay-per-token

1. Sign up and confirm your email.
2. Create a key on the API keys↗ page. The full key is shown once.
3. Set a daily and monthly dollar limit on the key so that a runaway loop cannot eat the balance.
4. Top up. SBP is charged in rubles at the market rate and the payment page adds an 8.5% fee on top: for $5 that is 5 × 1.085 = $5.425 in conversion. Crypto and MasterCard cards (fee about 13%, amounts $2 to $594) are also available.
5. Send a request to https://api.zukko.pro/v1 with the model gemini-3.8-flash.

from openai import OpenAI

client = OpenAI(base_url="https://api.zukko.pro/v1", api_key="fetch_your_key")
r = client.chat.completions.create(
    model="gemini-3.8-flash",
    messages=[{"role": "user", "content": "Answer in one word: connection test"}],
)
print(r.choices[0].message.content)

Models and prices: the models page↗, plus GET /v1/models. If you are just getting to know APIs, start with the article on what you can get through an API without paying↗.

FAQ

Is the Gemini API completely free?

No. Google has a free tier with limited quotas and a condition that data may be used to improve products, while serious work runs on the paid tier.

What request limits does the free tier have?

Google names three dimensions: RPM, TPM and RPD. Exact numbers depend on the model and project and are shown in AI Studio. They are not guaranteed and may change.

Can I use the free tier from Russia?

According to Google's available regions page, Russia is not in the list. If you need Gemini from Russia, use a gateway and pay by SBP, MasterCard or crypto.

What if I run out of limit?

Wait for the reset (the daily limit resets at midnight Pacific time), lower the request rate, or move to pay-per-token. Retry a 429 error with a growing pause.