Gemini API: Kostenloser Tarif bleibt 2026 bestehen, wird aber schmaler

Posted on Oct 23 • Originally published at apidog.com

The Gemini API free tier still exists in October 2026, but it's narrower than it used to be. You can call Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, Gemma 4 and several audio and embedding models for $0, with no billing account. Gemini 4 Argon, the image models and the video models are paid only. Google no longer prints per-model RPM and RPD numbers in its docs, and you're only billed after you link a billing account and prepay at least $5.

Every fact below comes from Google's pricing page, rate limits page, billing guide and Gemini API terms, read on October 22, 2026. Still need a key? Our guide to getting a free Gemini API key covers that step.

Yes, partly. The Gemini API free tier gives every Google AI Studio project free input and output tokens on selected models, with no card and no billing account. Google describes it as "Limited access to certain models." In exchange, Google can use your prompts and responses to improve its products, and your rate limits are lower than on paid tiers.

Here's the Free Tier column for Standard requests on October 22, 2026, with paid prices per 1M tokens.

Audio fares better: Gemini 3.8 Live, 3.8 Flash TTS, 3.5 Transcribe and Gemini Embedding 2 all list free input and output.

Argon gets its own note on the pricing page: "Gemini 4 Argon is not available on the Free tier." Our Gemini 4 Argon pricing breakdown covers its paid rates and the December 1 price change.

Watch the 2.5 models too. The models page says Google is "limiting access to the 2.5 models to users who have actively used them in the past." Gemini 2.5 Pro and 2.5 Flash still show "Free of charge," but a new project may not reach them. Google points new work to 3.5 Flash-Lite or 3.8 Flash.

Google publishes the rules, not the numbers. The rate limits page no longer carries a per-model RPM, TPM and RPD table. It sends you to the rate-limit view in Google AI Studio, which shows the active limits for your own project.

Check your real numbers in AI Studio before you design a retry policy. When you hit a ceiling, the API answers 429 RESOURCE_EXHAUSTED; our guide to using Gemini 3.8 Flash for free shows backoff for the Flash models.

The free tier is free in dollars, not in data. The terms treat "unpaid quota on Gemini API" as an Unpaid Service, and for those Google "uses the content you submit to the Services and any generated responses to provide, improve, and develop Google products." They add: "human reviewers may read, annotate, and process your API input and output."

So keep customer records, secrets and anything under NDA off a free-tier key.

Two regional rules apply in the European Economic Area, Switzerland and the United Kingdom:

You're billed only after you upgrade a project. A free-tier project has no billing account, so nothing on it turns into a charge. The billing guide lays out the path:

A clean setup is two projects: a free one for experiments and a paid one with a project spend cap for real traffic.

With a key from a free-tier project, this REST call to Gemini 3.8 Flash costs $0:

export GEMINI_API_KEY = "your-free-tier-key" curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.8-flash:generateContent" \ -H "x-goog-[REDACTED CREDENTIAL] \ -H " Content-Type: application/json " \ -X POST \ -d '{" contents ": [{" parts ": [{" text ": " Explain HTTP 429 in one sentence. "}]}]}' Enter fullscreen mode Exit fullscreen mode A 200 with text in candidates.content.parts.text means the free tier served you. Don't expect the same key to answer for gemini-4-argon: Argon needs a paid project. Keep the key in an environment variable, never in a commit.

Skip checking model IDs and limits by hand. With Apidog CLI, Claude Code, Codex, Cursor or OpenCode can write a test case for the Gemini endpoint, run it against your free key and report any 429.

npm install -g apidog-cli@latest apidog login --with-token <YOUR_TOKEN> npx -y skills add https://apidog.com # pick your agent when prompted Enter fullscreen mode Exit fullscreen mode Then give your agent one [REDACTED PROMPT] In Apidog project, create an environment "gemini-free" with base URL https://generativelanguage.googleapis.com. Add a test case for POST /v1beta/models/gemini-3.8-flash:generateContent with header x-goog-[REDACTED CREDENTIAL] that asserts status 200 and non-empty candidates.content.parts.text. Run it 5 times, 1 second apart, and tell me if any run returned 429.

The agent reads the payload schema, validates its JSON, then creates and runs the case: Enter fullscreen mode Exit fullscreen mode bash apidog environment create gemini-free --project --base-url https://generativelanguage.googleapis.com apidog cli-schema get test-case-create apidog test-case category --project apidog cli-schema validate test-case-create --file case.json apidog test-case create --project --file case.json apidog run --test-case --project -e \ --env-var "GEMINI_API_KEY=$GEMINI_API_KEY" -n 5 --delay-request 1000 -r cli,json