Gemini API billing: how to pay for Gemini by the token from Russia
How Google bills the Gemini API, where setup gets stuck on payment, and how to call Gemini from a single dollar balance instead.
6 min · · payments · models
If you want to add Gemini to your app, the first thing you hit is Gemini API billing: a Google project, a billing account, prepay and tier limits. For many developers in Russia this step is where everything stops. Below we go through how Google billing works, what matters in it, and how to pay for Gemini per token from a gateway balance without opening a Google billing account.
How Gemini API billing works at Google
What follows is a summary of Google's official Gemini API billing page as it looked on October 18, 2026. The page changes, so verify numbers against the original.
- New accounts start on the Free Tier with rate limits. To leave it, you link a Cloud Billing account to the project.
- Your tier depends on payment history: Tier 1 needs a linked billing account, Tier 2 and Tier 3 need accumulated payments and time since the first payment. Each tier has a monthly spend cap at the billing-account level.
- There are two plans: Prepay and Postpay. Google is moving developers to Prepay: the minimum credit purchase is $5, and unused credits expire after 12 months and are not refunded.
- When the prepaid balance reaches zero, every key in every project linked to that billing account stops working. Requests get HTTP 402 Payment Required.
- Costs are recorded with a delay of about ten minutes, so long jobs can overshoot the balance or the project cap.
- Google Cloud welcome credits for accounts opened after March 2, 2026 cannot be spent on the Gemini API.
Note the general logic: an API key has no billing settings of its own, it inherits the status of its project, and the project inherits it from the billing account. That suits an organization, but for a solo developer it means that without a working billing account there are no paid features.
Why billing is hard to set up from Russia
The paid tier needs a Google billing account in a supported country and a payment method that Google accepts. The list of methods and currencies is in the Cloud Billing section "Supported currencies and payment methods". If you have no accepted method, billing setup does not finish and paid models stay unavailable. This is not a skill issue: project settings cannot fix it.
We do not recommend picking a "convenient" registration country or using someone else's payment details: an account with inconsistent data can be restricted together with your keys. The legitimate paths are to pay Google with a method it accepts, or to reach Gemini through a gateway where payment is separate. For Russia the wording is: access from Russia through the gateway, payment by SBP, MasterCard or crypto.
Can you "buy" a Gemini API key
Queries such as "buy gemini api key" are understandable: people want a working key without account hassle. But at Google a key is created for free and you pay for usage. So a seller of keys is actually selling access to someone else's billing or someone else's project. Such a key can be revoked at any moment, and you do not control what happens to your requests. It is safer to get your own key at a gateway where the balance is yours and spending is visible in the dashboard.
Gemini per token from a gateway balance
Zukko has one fetch_... key and one dollar balance. Gemini models are available in the OpenAI-compatible format. The base URL for OpenAI clients is https://api.zukko.pro/v1. If by "buy Gemini API" you mean "start paying per token", the order is:
- Top up the balance in the dashboard (from $1) and create a key on the API keys page.
- Set the gateway
base_urland the modelgemini-3.8-flashorgemini-3.7-flashin your client. - Send a request.
from openai import OpenAI
client = OpenAI(
base_url="https://api.zukko.pro/v1",
api_key="fetch_your_key",
)
response = client.chat.completions.create(
model="gemini-3.8-flash",
messages=[{"role": "user", "content": "Summarize briefly: ..."}],
)
print(response.choices[0].message.content)
print(response.usage.prompt_tokens, response.usage.completion_tokens)The current list of models and prices comes from GET https://api.zukko.pro/v1/models and from the "Models" page in the dashboard. More on Python clients: Python API reference.
What it costs: the math
Prices are per 1M tokens, as of October 18, 2026.
| Model | Zukko in / out | Official in / out | Context | Strength index |
|---|---|---|---|---|
gemini-3.8-flash | $0.125 / $0.6 | $0.75 / $3.75 | 1M | 41 |
gemini-3.7-flash | $0.125 / $0.6 | $0.75 / $3.75 | 1M | 38 |
Example: in a month an app sends 10M input and 2M output tokens.
- Through Zukko: 10 × 0.125 + 2 × 0.6 = 1.25 + 1.20 = $2.45.
- By the official table: 10 × 0.75 + 2 × 3.75 = 7.50 + 7.50 = $15.00.
Cache is billed separately: a cache read costs $0.013 per 1M for these models and is counted when the provider reports it in the response; cache writes are not charged separately. For your own volumes use the formula: fresh input × input price + cache read × read price + output × output price. The strength index comes from Artificial Analysis; check quality on your own task.
What you give up compared with direct billing
Honest limits:
- For
gemini-3.8-flashandgemini-3.7-flash, image input is marked "no" in the Zukko table. If you need images, pick a model marked "yes" in the list. - Through the gateway you work with the OpenAI or Anthropic format. Features specific to the native Google SDK (some AI Studio parameters) may be unavailable. Test your scenario with a sample request.
- Instead of Google's caps you get Zukko limits: every key has a daily and monthly limit in dollars and a list of allowed models (
allowed_models). More: key limits and cache.
Errors are transparent: 402 means the balance is too low, 403 model_not_allowed_for_key means the model is not allowed for the key, and 429 means the key limit is used up or the model provider is limiting you.
How to top up
- SBP: payment in rubles at the market rate at the time of payment. The payment page adds an 8.5% fee on top, and the balance is credited with the dollar amount you requested. For example, for a $10 top-up the page charges 10 × 1.085 = $10.85 in rubles.
- Cryptocurrency (USDT and others).
- MasterCard cards: a fee of about 13%, amounts from $2 to $594.
Top-ups start from $1 and there is no subscription: money is spent only on tokens. More about payments in the article on paying via SBP.
FAQ
Do I need a Google Cloud account to call Gemini through Zukko?
No. Requests go to api.zukko.pro with a fetch_... key, and no Google billing account is involved. You see the balance and spending in the Zukko dashboard.
What happens when the balance runs out?
Requests start returning 402. This resembles Google's behavior at zero prepay, but the balance is yours and a top-up takes minutes. A daily limit on the key protects you from surprise spending.
How do gemini-3.8-flash and gemini-3.7-flash differ?
By the table they cost the same, 3.8 has a strength index of 41 and 3.7 has 38. Both have a 1M context. Compare both on your own prompts: an index gap does not guarantee a gap on your task.
Can I call Gemini in the Anthropic format?
Any model can be called in either format, OpenAI or Anthropic. For Gemini the OpenAI format is simpler: base https://api.zukko.pro/v1.
To try it, create a key in the dashboard and follow the quick start: the first request takes a few minutes.
zukko
Pay 25×
less
for Claude Code & ChatGPT
One fetch_* key and base URL api.zukko.pro — for Cursor, Claude Code, Codex, and any SDK.
Related articles
- xAI Grok API: Connecting Grok 4.7, Price and the 256K ContextHow to call grok-4.7 through an OpenAI-compatible API, what to do about the 256K context limit and what typical requests cost. Plus a comparison with Grok 4.6.
- GPT-6 Astra or Claude Fable 5.1: Two Top Models by Price and IndexTwo premium models with the same intelligence index and different Zukko prices. We price a single request and an agent session with cache, and explain how to choose on your own data.
- Qwen or DeepSeek via API: Price, Context and Intelligence IndexFour models, one task and an honest calculation: what a run costs on Qwen 3.8 Max and Flash versus DeepSeek V4 Pro and V4.1 Flash.
- Prompt Engineering Examples: 12 Ready-Made Templates for WorkTwelve prompt templates for everyday tasks: what each one says, why it is built that way, and what a single API call costs.