Minute Verdict
Google AI Studio is the browser workbench Google runs for its Gemini models: a prompt editor, a model picker, API key management and an app gallery. The workbench itself is free in every region where it is available, and nothing is billed until a project is linked to a billing account. After that the meter is per token rather than per seat, from $0.10 per million input tokens on Gemini 2.5 Flash-Lite.
Best for: a developer who wants to test prompts and models before wiring them into an application. Not for: a buyer who wants a finished assistant to talk to, since that is the Gemini app, and AI Studio hands you parameters instead. From: $0.10 per million input tokens, checked November 3, 2026
What Google AI Studio does
One web surface, three jobs. A prompt box, a model selector and the settings that go with them, so a prompt can be tried against several Gemini models without leaving the page. API key management, since the same account issues and monitors the keys an application will use. And a gallery of sample projects that can be remixed.
On the day of this review the pricing page carried Gemini 4 Argon as the flagship reasoning model, Gemini 3.8 Flash and 3.6 Flash for general work, Gemini 3.5 and 3.1 Flash-Lite for volume, Nano Banana 2.1 for images, Gemini Omni Flash for video, three text-to-speech models, a family of Live models for real-time audio, and Lyria for music. One account reaches text, image, video, speech and music.
Around the models sit the developer tools: official SDKs for Python, JavaScript, Go, Java and C++, an Interactions API for stateful agents, and function calling with native Search and Maps grounding.
Two comparisons settle what the product is. It is not the Gemini app, which is the consumer chat: AI Studio exposes the model, its parameters and its key. It is not Vertex AI either, the enterprise platform the same page points buyers to for support, security and provisioned throughput, with a default quota of 1,000,000 search queries per day there against 1,500 per day on the developer API.
On this page
What you get for the money
| Plan | Price | What is included | The limit that matters |
|---|---|---|---|
| Free | No charge | Limited access to certain models, free input and output tokens, AI Studio access | Content is used to improve Google products, and Gemini 4 Argon is not available |
| Paid | From $0.10 per 1M input tokens | Higher rate limits, context caching, Batch API at 50% off, access to the most advanced models | A billing account is required, and spend is capped per usage tier |
| Enterprise | Contact sales | Dedicated support, advanced security and compliance, provisioned throughput | Sold through Gemini Enterprise Agent Platform, not from the studio |
The free tier is not a trial credit. It carries input and output tokens on the models open to it, which is how a solo developer evaluates Gemini without a card. The paid tier raises the rate limits, opens context caching and the Batch API, and makes the strongest models selectable. Batch work is billed at half the standard rate, the largest lever on a large job.
The per-token rates for the models that matter most read as follows, in dollars per million tokens.
| Model | Input | Output | Free tier |
|---|---|---|---|
| Gemini 4 Argon | $2.00 until November 30, 2026, then $4.00 | $10.00, then $20.00 | No |
| Gemini 3.8 Flash | $0.75 until December 31, 2026, then $1.50 | $3.75, then $7.50 | Yes |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | Yes |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | Yes |
| Nano Banana 2.1 | $1.50, images at $30.00 | $7.50, or $0.0336 per 1K image | No |
| Gemini Omni Flash | $1.50 | $17.50 for video, about $0.10 per second at 720p | No |
There is no twelve-month total to quote, because the bill follows volume rather than a subscription. Two charges sit outside the tables. Grounding with Google Search gives 5,000 free search requests a month across Gemini 3 and newer models, then bills $14 per 1,000, and the developer API caps that at 1,500 requests per day. Image output is metered in tokens too, so a Nano Banana 2.1 image at 4K costs $0.113 in output alone, thirty-three times the same image at 1K.
Where it falls short
The free tier is paid for with data. The API terms state that on unpaid services, AI Studio and the unpaid Gemini API quota included, Google uses what you submit and what the model returns to provide, improve and develop its products, and that human reviewers may read, annotate and process input and output. The document adds that the data is disconnected from the account, the API key and the project before reviewers see it. On the paid quota, prompts and responses are not used to improve products and are handled under a data processing addendum. For a buyer handling client material, that sentence decides the tier, not the price.
The flagship model is missing from the free tier. Gemini 4 Argon is marked as not available there, and the free tier is described as limited access to certain models, so the model a reader most wants to try is the one that requires billing.
Two increases are already dated. Gemini 4 Argon goes from $2.00 to $4.00 per million input tokens and from $10.00 to $20.00 on output on December 1, 2026. Gemini 3.8 Flash goes from $0.75 to $1.50 and from $3.75 to $7.50 on January 1, 2027. A budget built on today’s rates is wrong in two months.
Model churn is a maintenance cost. Gemini 2.5 Pro and Gemini 2.5 Flash can no longer be selected in AI Studio, and requests to them route to Gemini 3.8 Flash and Gemini 3.5 Flash-Lite from December 1, 2026. Gemini 2.5 Flash Image shuts down on March 15, 2027. A pinned model name expires.
Finally, the guard rails on a new billing account are tight. Spend limits run on a rolling ten-minute window: $10 at Tier 1, $200 at Tier 2 and Tier 3. A fast batch job can return a 429 error before it has spent what you expected, and the documented fix is to slow down or request an increase.
Google AI Studio alternatives
These are the three tools buyers compare against. Each is already reviewed here.
- DeepSeek, a chat that is free with an account and an API billed per token from $0.003 per million input tokens, the cheapest entry we track: the DeepSeek review.
- Emergent, a prompt-to-app builder for people who want a working application rather than a model endpoint, from $20 a month: the Emergent review.
- Grok, free with limits and $8 a month inside X Premium, aimed at chat rather than at an API key: the Grok review.
For document work rather than model work, NotebookLM is free and sits in the same Google account: the NotebookLM review covers its paid tiers. The category page for AI app builders lists the rest.
Who it is for: Google AI Studio
It fits a developer or a technical founder who has decided to build on Gemini and wants to test prompts, compare models and issue a key without a procurement cycle. The free tier covers evaluation honestly, and the jump to paid is a billing toggle rather than a sales call. It also fits a team already inside Google Cloud, since the project, key and quotas carry over when the prototype has to scale.
It does not fit a non-technical buyer who wants an assistant: no document workspace, no team chat, and the studio assumes you know what a token, a temperature and a system instruction are. It does not fit an enterprise that needs a contract and a support channel on day one, since those come through the enterprise platform. And it does not fit anyone who must keep prompts out of Google’s improvement loop while staying on the free tier, because that is what the free tier reserves.
How we checked
This is a research review. We have not used Google AI Studio, so there is no hands-on score, no latency measurement and no sample output.
Every price and limit here was read on November 3, 2026, on the Gemini developer pricing page, the rate limits documentation and the API terms, plus the studio home page for the feature list. The dated increases and the deprecation dates are quoted from those pages, not estimated. A test would settle what no document can: how the models behave on real prompts. There is no affiliate programme to disclose, and our editorial policy covers how the site is funded.
What to do next
Run one prompt you already know the answer to on Gemini 3.8 Flash before you attach a card. If the output is good enough, run the same prompt on 2.5 Flash-Lite and on Gemini 4 Argon and compare the gap with the price gap. If you build, put bulk work through the Batch API at half rate, keep prompts stable so caching does the work, and rebuild your budget for December 1, 2026 and January 1, 2027.
FAQ
Is Google AI Studio free? Yes. The pricing page states that AI Studio usage is free of charge in all available regions, and the free tier includes input and output tokens with limited access to certain models.
How much does the Gemini API cost? From $0.10 per million input tokens and $0.40 per million output tokens on Gemini 2.5 Flash-Lite. Gemini 3.5 Flash-Lite is $0.30 and $2.50, Gemini 3.8 Flash is $0.75 and $3.75, and Gemini 4 Argon is $2.00 and $10.00 until November 30, 2026.
Does Google train on my prompts in AI Studio? On the unpaid services, including AI Studio and the unpaid Gemini API quota, Google states that content is used to improve its products and that human reviewers may read input and output. On the paid quota it states that prompts and responses are not used to improve its products.
Can I build an app in Google AI Studio? The home page advertises vibe coding, an app gallery and remixable projects, alongside API key management and the SDKs. The API is the surface to plan for once the prototype has to serve users.
Do the prices change soon? Yes. Gemini 4 Argon moves from $2.00 to $4.00 per million input tokens on December 1, 2026, and Gemini 3.8 Flash moves from $0.75 to $1.50 on January 1, 2027.
Sources
Sources
- https://ai.google.dev/pricing (checked November 3, 2026)
- https://ai.google.dev/gemini-api/docs/rate-limits (checked November 3, 2026)
- https://ai.google.dev/gemini-api/terms (checked November 3, 2026)
- https://aistudio.google.com/ (checked November 3, 2026)
Update log
Published November 3, 2026. Prices and limits checked November 3, 2026 against the Gemini developer pricing page, the rate limits documentation and the API terms.
