Free Google AI: the 7 platforms worth your time
Google's free AI goes beyond Gemini: discover the 7 platforms you can use without paying, what each does best, and exactly where the free tier ends.
NinaKalkazu.org

You open your phone, type a question into Gemini, get the answer and close the app. Done. For most people, that's the entire Google AI experience.
Except it's not. Behind that text field sits a collection of platforms you already have the right to use now, without a credit card and without a trial period: a studio that runs the models directly, a study assistant that only answers based on your files, an image generator and three more tools for programmers.
The problem is that nobody tells you where the free tier ends. This article does both things: it shows the seven platforms worth your time and marks, in each one, where Google starts charging — including cases where the company simply doesn't disclose the limit. Every price and every quota here came from the company's official pages, read on September 19, 2026; usage reports come from the articles cited throughout the text.
Quick summary: the answer in 30 seconds. With a regular Google account you can use the Gemini app, Gemini Notebook (formerly NotebookLM), Google AI Studio, the Nano Banana image generator, Gemini Code Assist, Gemini CLI and Antigravity, without paying. What money buys isn't access: it's volume and priority. Paid plans start at R$ 24.99 per month (in Brazilian reais), and students get the R$ 96.99 per month plan free for one year.
Index
- Why Google's free AI is the most generous available today
- Gemini: the assistant already on your phone
- Gemini Notebook: formerly NotebookLM, and why it makes fewer mistakes
- Google AI Studio: the models without an intermediary
- Nano Banana: free images inside Gemini
- The free tools for programmers
- Where the free tier ends, and how much it costs to cross the line
- Which one to use in each situation
- The price that doesn't appear on the bill
- Frequently asked questions
Why Google's free AI is the most generous available today
Because Google isn't selling AI to you. It's buying your habit. Each model run for free costs the company money, and it pays that bill so you don't develop the habit of opening a competitor's product. It's the same logic Gmail and Google Docs used two decades ago.
This has a practical consequence: the free tier boundary moves. A feature released today might become a subscription item next month, and the opposite also happens.
Before continuing, a distinction that avoids confusion. The model is the brain, the trained program that generates text — in Google's case, the Gemini family. The platform is the door through which you talk to it: an app, a website, a terminal.
One platform can offer multiple models, and the same model appears on multiple platforms. This matters because the free plan limit is almost always on the platform, not the model.
A Google account is the full ticket
None of the seven tools in this article asks for a credit card to start. You log in with the same account as Gmail and you're in. The official Gemini Code Assist page is explicit in saying that the individual version costs US$ 0 per month, with no card at all.
This detail separates Google from much of the competition, where "free" often means a few days of trial with a card registered beforehand.
In one sentence: Google's free tier is commercial strategy, not kindness, and that's why it's both large and unstable at the same time.
Gemini: the assistant already on your phone
Gemini is Google's general-use chat, the direct equivalent to ChatGPT. It's on the website, on the Android and iPhone app, and appears increasingly inside Search, Gmail and Docs. To ask a question, summarize a document or draft a text, this is where the conversation starts.
A change at the end of 2025 significantly improved the free version. Google switched the app's default model to Gemini 3 Flash, which, according to the announcement reported by The Verge, thinks like the Pro model but responds at the speed of a lightweight model. Tulsee Doshi, product director at Google DeepMind, told the site that the switch brings faster and more detailed responses than the previous model.
In practice, the non-paying user stopped receiving leftovers.
What the free plan limits isn't the quality of the response, but the quantity. More expensive resources to run, like long searches and video generation, appear with smaller quotas or are reserved for subscribers. Google's plans page describes the paid tiers as "more access" to the same Gemini, not as access to a different Gemini.
Where exactly the free app ceiling is, however, Google doesn't say. There's no official page with the number of questions per day, and this limit changes according to current demand. It's the first of the invisible boundaries in this article.
Gemini Notebook: formerly NotebookLM, and why it makes fewer mistakes
If you're only going to try one tool from this list, make it this one. Gemini Notebook is a research assistant that only answers based on the files you uploaded: PDFs, web pages, YouTube videos, pasted text. It reads that and nothing beyond that.
This restriction is the beauty of the thing. The technical term is source-grounded: instead of pulling content from training memory, the model is forced to answer by looking at the material you provided, and to cite where it got each statement.
Think of the difference between asking a friend who has read a lot about a subject and asking an intern who just read exactly your ten documents. The friend mixes things up; the intern points to the page. Google states, on the product's plans page, that this grounding reduces the chance of the system inventing information.
For readers in Brazil, this solves a concrete problem: studying for competitive exams, understanding a contract, reviewing course readings. You upload the material and talk to it. Notebook also transforms everything into an audio summary in podcast format, mind map, timeline, flashcards or presentation.
What changes with the new name
In July 2026, NotebookLM became Gemini Notebook. The change is name only: the official page confirms it's the same product and that old notebooks are all still there.
Before that, in April, Google had already embedded Notebook inside the Gemini app itself, as Engadget reported: you can create a notebook without leaving the chat. The rollout started with subscribers and reached free users later.
Up to 50 sources per notebook
That's the number that matters. On the free plan you have up to 50 sources per notebook, according to the official plans page checked on 09/19/2026. How many questions and how many audios fit per day, Google doesn't say: it calls these ceilings just "standard limits."
What you can know is how much money multiplies that standard. AI Plus doubles usage limits and goes up to 100 sources per notebook. AI Pro quadruples and goes to 300 sources. AI Ultra accepts 600 sources and reaches twenty times the Pro limits.
Fifty sources is quite a lot. You can throw an entire semester of readings into a single notebook. Who exceeds this ceiling is usually a researcher doing literature review.
Google AI Studio: the models without an intermediary
AI Studio is the workbench Google created for developers and ended up becoming the best entry point for the curious. You open the site, choose which Gemini model you want to use, adjust its behavior and chat. No app, no installation, no card.
The difference from the Gemini app is control. In the app, Google decides the model and rules for you. In AI Studio, you choose among available versions, adjust settings like response length and how much the model can invent, and even see the text the system uses as secret instruction.
There's a second, more ambitious use: describe a program in Portuguese and watch AI Studio build it. Sean Hollister, from The Verge, made three Android apps in one afternoon by typing 148 words in the browser, the length of three paragraphs of this article. His account brings the honest counterbalance: the apps worked, but were poor, and he hit the daily limit of the free plan right when he started improving them.
API key without a card
Here's the part almost nobody takes advantage of. AI Studio distributes an API key — the password that allows your program to talk to Gemini automatically — and there's a Free tier for this API.
The official Gemini API documentation, on the usage limits page updated September 2, 2026, confirms that the free tier depends only on an active project, with no billing account linked. Paid tiers begin when you register billing.
Two caveats that the same page makes clear. First: limits are measured in requests per minute, tokens per minute and requests per day, and vary by model. The token is the piece of word the model processes at a time, something like AI's syllable — that's what the bill is calculated by.
The second caveat: Google warns that published limits are not guaranteed and may change according to server capacity.
For a personal project, a WhatsApp bot or a spreadsheet that summarizes itself, this tier usually works. For something that goes live with real users, it doesn't.
Nano Banana: free images inside Gemini
Google's image generator goes by the nickname Nano Banana, and the Pro version arrived with Gemini 3. It does what you'd expect from a current generator, with an unusual specialty: writing legible text inside the image. This makes the tool useful for posters, infographics and diagrams, not just pretty illustrations.
The tool is free with a daily quota in the Gemini app, and the quota increases for subscribers, according to Engadget's report at launch. This is the second invisible boundary: Google confirms there's a daily ceiling, but doesn't publish how many images, and the number changes over time.
A detail worth noting: images come out with a hidden watermark inside the file, in the industry standard called C2PA, which identifies it as AI-generated. If you plan to use the image commercially, know that it carries this mark.
The free tools for programmers
There are three platforms on this list aimed at code, all free for individuals: Gemini Code Assist, Gemini CLI and Antigravity. In two of them Google publishes the limit with a number, which is rare in the rest of this article. In Antigravity, it doesn't.
Gemini Code Assist lives inside your editor and completes code as you write. The official page states, for the individual version, 6,000 code-related requests and 240 chat conversations per day, in VS Code, JetBrains and Android Studio. It's a high ceiling for someone programming alone: who gets close to it is usually someone who automated their own workflow.
It also reviews code changes proposed by you or colleagues on GitHub — the so-called pull requests — pointing out bugs and suggesting fixes.
Gemini CLI brings the same assistant to the terminal, that black screen where programmers type commands instead of clicking. The free tier gives 1,000 model requests per day. The cheapest business license, at US$ 19 per user per month, gives 1,500, and the US$ 45 one gives 2,000, according to the same page. In other words: without paying anything you're at two-thirds of the corporate team.
Antigravity is the most recent bet: an editor built around agents, AI programs that execute multi-step tasks on their own, instead of just suggesting the next line. If the idea is new to you, the concept is explained calmly in our article about what agentic AI is. It's the third invisible boundary on the list: Google offers the editor without charging, but doesn't publish the usage quota.
Outside the count of seven, because it's infrastructure and not an AI tool, there's still the Cloud Shell Editor: a programmer's computer already set up, that opens in a browser tab. It's 50 hours per week of machine time at no cost, something like seven hours a day, for those who don't want to install anything on their own machine.
Here's the caveat: the free version of Code Assist currently runs Gemini 2.5, and access to Gemini 3 is on a waiting list, released immediately only to those who subscribe to the Ultra plan. The volume is generous; the newest model is not.
Where the free tier ends, and how much it costs to cross the line
Crossing the line costs R$ 24.99 per month on the cheapest plan and R$ 96.99 on the main one. The general rule is simple: the free plan gives access to almost everything, but in smaller quantity and with the standard model. What the subscription buys is volume, queue priority and the most expensive resources to run, especially video generation and long, automated searches.
The prices below came from Google's Brazilian plans page, checked on 09/19/2026.
| Plan | Price in Brazil | Storage | What really changes |
|---|---|---|---|
| Free | R$ 0 | your account's | Standard model and smaller quotas on everything |
| Google AI Plus | R$ 24.99/month | 400 GB | Double Gemini access; 100 sources in Notebook |
| Google AI Pro | R$ 96.99/month | 5 TB | Quadruple access; 300 sources; higher limits in AI Studio and Antigravity |
| Google AI Ultra | from R$ 779.90/month | 20 TB | Up to 20 times Pro access; Deep Think reasoning mode; 600 sources |
The intermediate plan deserves a comment. AI Plus was born in emerging markets and only later reached the whole world, as TechCrunch reported in January 2026. It exists precisely to catch those who find the main plan expensive, and in Brazil it costs less than a third of the AI Pro price.
Students pay zero for one year
If you're a student, stop here and check this out. Google offers AI Pro free for one year for students — the R$ 96.99 per month plan — with the higher limits of Gemini, Notebook and YouTube Premium Lite included. The offer is announced on the plans page itself.
It's the best deal on this entire list, and also the most ignored. Twelve months of the plan equals over a thousand reais.
Which one to use in each situation
Too many tools paralyze. This table solves the choice by the task you have in front of you, not by the product name.
| If you want to… | Use | Why |
|---|---|---|
| Answer a quick question | Gemini app | It's on your phone and answers immediately |
| Study specific material | Gemini Notebook | Answers only from your sources and cites the page |
| Understand a contract or notice | Gemini Notebook | Doesn't mix with outside information |
| Test and compare models | Google AI Studio | You choose the model and settings |
| Build an app without knowing how to program | Google AI Studio | Describe in Portuguese and it builds |
| Automate something with your code | Google AI Studio | The API key has a free tier, no billing |
| Create poster, infographic or cover | Nano Banana in Gemini | Writes legible text in the image |
| Program day to day | Gemini Code Assist | 6,000 requests per day, no card |
| Work directly in the terminal | Gemini CLI | 1,000 requests per day, reads and executes files |
| Delegate an entire code task | Antigravity | Agents do multiple steps without supervision at each line |
The price that doesn't appear on the bill
Free in money isn't free in data. Your conversations with Gemini can be reviewed by people and used to improve the models, and Google itself writes this on the company's privacy pages.
Two concrete things to do today. First: turn off Gemini app activity in account settings, which prevents future conversations from being used to train the system. Second: never paste confidential documents, client data or passwords into any company's AI chat.
Gemini Notebook has a more protective policy than the regular chat. The official page states that, for individual use, your data is not used for training, except when you voluntarily send feedback about a response.
In one sentence: treat the regular chat as conversation at a bar table and Notebook as a locked drawer, which only opens if you send feedback yourself.
Frequently asked questions
Is Google AI really free or is it a trial?
It's truly free, with no expiration date and no credit card. What exists are usage quotas: you make a limited number of requests per day on each platform. When the quota runs out, you wait for the day to turn or subscribe to a plan.
What is the best free Google artificial intelligence for studying?
Gemini Notebook, by far. It only answers based on the files you uploaded, cites where it got each piece of information and transforms the material into audio summary, mind map and flashcards. The free plan accepts up to 50 sources per notebook.
Does free Gemini use a worse model?
Today, no. Since the end of 2025, Gemini 3 Flash has been the app's default model for everyone, paying or not. The difference between free and paid is much more in the number of requests per day and access to expensive features, like video, than in text quality.
Is NotebookLM gone?
No, it just changed names. In July 2026 it became Gemini Notebook, with the same notebooks and the same functions. The official page confirms that nothing was lost in the transition, and the product continues working as a separate application on the web and mobile.
Can you use Google's free artificial intelligence for work?
Yes, with two cautions. Check the numbers and names the AI generates, because they all make mistakes. And don't upload confidential company material on a personal account, since the privacy rules for a regular account are different from those of a corporate account.
Do I need to pay to generate images?
No. Nano Banana works in the Gemini app without a subscription, with a daily image quota. Subscribers receive a larger quota. Generated images carry an invisible seal that identifies them as AI-made.
If you use Google every day and have never gone past Gemini's text field, you're failing to use what's already yours. Start with Notebook this week, with three files you've been putting off reading, and ask for an audio summary.
In one afternoon you'll knock out the overdue reading and understand what these tools do differently from regular chat. To situate them in the larger terrain, it's worth browsing through the other texts in our artificial intelligence section.






