Google AI Studio

Google's free browser workbench for testing Gemini models and prompts before wiring them into a paid API integration.

Assistants & chat Model

Checked 25 Sep 2026

What you'd use it for

Working out which Gemini model and prompt wording actually does the job before a developer wires it into something that runs unattended.

Test a prompt against a Gemini model with adjustable temperature, system instructions and safety settings, no code required
Upload a PDF, image, audio or video file into the chat and ask questions about it directly
Get the equivalent Python, JavaScript, Go or REST code for a working prompt, ready to paste into a real build
Switch between Gemini model sizes on the same prompt to see which one is actually good enough

The verdict

Google AI Studio is the fastest way to find out what a Gemini model can actually do with your prompt, for free, before anyone builds on it. It rewards flipping between Gemini model sizes and settings in one tab, but it has no memory, no file storage, and does nothing on a schedule — every result has to be read and carried out by hand. The real cost only shows up once the prompt you tuned here gets wired into the Gemini API, where per-token pricing and whichever model size you picked casually while experimenting now determine the bill. Good for settling a model or prompt question fast; not a place to run an actual business process.

What it costs

Tier	Price	For	Includes
Free	$0	Developers and prompters trying models before paying for anything.	Full access to Google AI Studio, and the Gemini API's free tier at lower rate limits — data from free-tier use may be used to improve Google's products.
Paid (Gemini API)	Pay-as-you-go, billed per token (example: Gemini 3.8 Flash charges $0.75 per million input tokens through 31 December 2026, rising to $1.50 from 1 January 2027; output is priced at roughly 5x the input rate)	Production use that has outgrown the free tier's rate limits.	Higher rate limits, context caching, a Batch API at a 50% discount, and data excluded from model training.
Enterprise	Custom pricing	Organisations needing dedicated support and compliance guarantees.	Dedicated support channels, advanced security and compliance options, provisioned throughput, volume-based discounts.

Read from Google AI Studio's own pricing page on 25 Sep 2026. Vendors change prices without notice.

What you need to start

A Google account
Access from one of the roughly 195 countries and territories Google lists as supported for the Gemini API

A Google account gets you a working AI Studio session and a free API key within minutes; moving past the free tier needs a linked Google Cloud billing account and a $5 minimum prepay.

Your first run Sign in with a Google account, type a prompt into the chat pane, and a Gemini answer appears in seconds — no API key needed until you want the code to take it elsewhere.

Where it breaks

AI Studio itself is always free in every region Google lists. The cost that actually matters is the Gemini API's per-token rate once a project outgrows the free tier's rate limits, and that rate differs by model and changes on a published schedule — the step-up on 1 January 2027 above is one example, not the only one.

Runs entirely in the browser at aistudio.google.com. There is no desktop app; production use happens through the Gemini API rather than the Studio interface itself.

First entry in this reference. There is no previous check to compare it against.

When not to use it

Not for anyone who wants a finished workflow or assistant that keeps running once the tab is closed — AI Studio does nothing on its own and remembers nothing between sessions. Also not the place to paste in sensitive customer data on the free tier, since Google may use that input to improve its models unless you move to a paid, training-excluded tier.

Google AI Studio is the tab you open to find out whether a Gemini model can actually do the thing you have in mind, before anyone writes a line of integration code. Type a prompt, attach a file, tune the settings, and see the raw output — no API key required to start, no bill attached to the experiment.

It is a browser tab where you read an answer and copy it out by hand, or paste in the next thing you want checked. It has no memory of its own across sessions in the way Claude or ChatGPT does, and no file access to your machine — everything it sees is what you paste or upload into that one conversation.

The real decision happens after the tab

Nobody pays for AI Studio. What a business actually budgets for is what comes next: building the same prompt into the Gemini API, at which point Google starts charging per token, and the rate depends on which model you picked while experimenting in the free workbench. A prompt that works cheaply on a smaller Gemini model in AI Studio can get expensive fast if the production build defaults to a larger one out of habit.

Where it fits

Treat AI Studio as the fitting room, not the store. It is where you settle model choice and prompt wording before a developer — or a tool like Make calling the API directly — wires the result into something that runs without you watching it.

Alternatives

Assistants & chatModel

Claude

Anthropic's AI assistant, available in a browser tab and — the part that matters here — as a desktop app that can read the files on your own machine and reach the systems you connect to it.

Checked 15 Sep 2026

Assistants & chatApps

ChatGPT

OpenAI's assistant in a browser tab, a desktop app and a phone app, where you ask questions, draft text, and hand it files or spreadsheets to work through.

Checked 25 Sep 2026

Assistants & chatApps

Google Gemini

Google's assistant, usable on its own and also inside Gmail, Docs, Sheets and Drive, so it can draft and summarise against files you already keep in Google.

Checked 25 Sep 2026