How to Get a Gemini API Key in 2026: The Free Tier, Billing, and Startup Credits

From Google's API key docs. The whole thing takes about a minute if you have never used Google Cloud before.

Google's SDKs read the key from the GEMINI_API_KEY or GOOGLE_API_KEY environment variable automatically. If both are set, GOOGLE_API_KEY wins. Set it once in your shell:

export GEMINI_API_KEY="YOUR_API_KEY"

Python

Install the SDK with pip install -U google-genai, then:

from google import genai

client = genai.Client()
interaction = client.interactions.create(
    model="gemini-3.8-flash",
    input="Explain how AI works in a few words"
)
print(interaction.output_text)

JavaScript

Install with npm install @google/genai, then:

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({});
const interaction = await ai.interactions.create({
  model: "gemini-3.8-flash",
  input: "Explain how AI works in a few words"
});
console.log(interaction.output_text);

curl

curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model": "gemini-3.8-flash", "input": "Explain how AI works in a few words"}'

These use Google's newer Interactions API, which is now the default in the quickstart. The older generate_content call still works if you are following an older tutorial.

"Free" means free input and output tokens, with rate limits instead of a bill. Model availability from Google's pricing page; paid rates are per million tokens.

Move to the paid tier when you hit the free limits, need Gemini 3.1 Pro or the image models, or are about to handle customer data. Paid requests are not used to improve Google's products.

Click Set up billing on the API keys or Projects page in AI Studio, link or create a Cloud Billing account, and add a payment method. New accounts are on Prepay: you buy credit up front, with a $5 minimum.

Three Prepay rules catch people out:

• At $0 your keys stop. Requests fail with a 402 error. The project does not drop back to the free tier. Turn on auto-reload if you are running anything in production.
• Prepaid credit expires after 12 months and is non-refundable.
• The $300 Google Cloud welcome credit does not work on the Gemini API for accounts opened after March 2, 2026.

Paid accounts move up usage tiers as they spend. Tier 1 is instant when you link billing and caps spend at $250 a month. Tier 2 unlocks after $100 paid and 3 days, with a $2,000 cap. Tier 3 unlocks after $1,000 paid and 30 days. Spend on any Google Cloud service counts toward the upgrade.

If you are in the UK, Switzerland or the EEA, Google's paid data terms apply even on the free tier, but you can only use the paid tier to serve an app to users there.

A leaked key on a billed project is a blank cheque. Google's own rules:

• Never commit it to Git. Keep it in an environment variable or a secrets manager such as Google Cloud Secret Manager.
• Never put it in a web or mobile app. Anyone can pull it out of the client. Call Gemini from your own backend instead.
• Restrict it. In AI Studio, any key marked "Unrestricted" can be limited to the Gemini API only. In the Cloud Console you can also restrict it to your server's IP addresses.
• Set a billing alert so a runaway script or leaked key shows up in hours, not at the end of the month.

If a key leaks, create a new one, deploy it, then disable the old key once the new one works, and check your usage. Google blocks keys it detects as leaked with the message "Your API key was reported as leaked." Unrestricted keys left unused for a long time can also be blocked, so create a fresh restricted key if an old one stops working.

From Google's troubleshooting guide. The SDKs retry 429 and 5xx errors automatically; do not retry 400, 402 or 403.

Google offers Gemini two ways. The Gemini Developer API is what you get with an AI Studio key, and it is what Google recommends for most developers. Vertex AI, which Google's docs now call the Gemini Enterprise Agent Platform, runs the same models inside Google Cloud with service-account authentication, regional controls and enterprise compliance.

Start with an AI Studio key. Move to Vertex AI when a customer asks for data residency or enterprise controls, or when your whole stack already runs on Google Cloud. The SDK is the same; you switch with one setting on the client.