Google AI Studio API limits: free quotas, and when to pay
Short answer The Gemini API free tier still exists, but Google publishes no fixed free numbers for most models: each project's limits show in AI Studio and can change with demand. The newest Pro-class, image and video models are paid only, and free prompts may be read by human reviewers. Free suits testing; customer-facing or daily work belongs on a paid key.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhat are the Google AI Studio API limits right now?
There is no single public number. Google's docs no longer list free-tier quotas for most models. Your project's real limits sit on the rate limit page inside AI Studio, and Google warns they can move:
"Specified rate limits are not guaranteed and actual capacity may vary." (Gemini API docs: Rate limits)
Every limit is counted three ways, and hitting any one of them stops your requests:
| Limit | What it counts | Resets |
|---|---|---|
| RPM (requests per minute) | Calls you send in one minute | Every minute |
| TPM (tokens per minute) | Input text sent in one minute; a token is roughly a short word | Every minute |
| RPD (requests per day) | Calls in one day | Midnight Pacific time |
Two rules catch most owners out. Limits apply "per project, not per API key" (Gemini API docs: Rate limits), so making a second key in the same project buys nothing. And the daily count resets at midnight Pacific, which is 3 a.m. in Toronto or New York.
What changed with the free tier?
The free tier has shrunk in steps, with little notice:
- December 2025: Gemini 2.5 Pro dropped off the free tier. Asked if it was temporary, a Google forum moderator replied that the change "applies to all users and does not appear to be temporary" (Google AI Developers Forum, December 10, 2025).
- January 2026: free use of the AI Studio chat screen was cut to daily caps in the tens of requests, users reported on Google's forum. A Google staff member answered that "sometimes we have to adjust rate limits based on demand" (Google AI Developers Forum, January 24, 2026).
- November 2026: the newest top models launched paid only. Gemini 4 Carbon, released November 10, "is not available on the Free tier" (Gemini API docs: Pricing).
- December 1, 2026: Gemini 2.5 Pro and 2.5 Flash are deprecated and already gone from AI Studio's model picker. From this date, API calls to them are routed to Gemini 3.8 Flash and Gemini 3.5 Flash-Lite, billed at those models' rates (Gemini API docs: Pricing).
The lesson: if a workflow you rely on was built for the free tier, assume its limits will change again.
Which Gemini models are still free?
Text and voice models mostly are; the most capable reasoning models and anything that makes images or video are not. From Google's pricing page as of November 16, 2026:
| Free tier | Paid only |
|---|---|
| Gemini 4 Hydrogen | Gemini 4 Carbon, Gemini 4 Argon |
| Gemini 3.8 Flash, 3.6 Flash | Gemini 3.1 Pro Preview |
| Gemini 3.5 and 3.1 Flash-Lite | Image models (Nano Banana 2.1, Nano Banana Pro, 3.1 Flash Image) |
| Speech and live voice models | Gemini Omni Flash (video) |
| Embedding models | Grounding with Google Search on Gemini 3 and newer |
That last row matters for marketing work. Grounding with Google Search lets the model check live search results before it answers. Without it, a free model writes from its training data, which can be out of date on prices, rules and local facts.
Is the free tier safe for customer data?
No. Google's terms say free use, including the AI Studio chat screen, may be used to improve Google products, and:
"To help with quality and improve our products, human reviewers may read, annotate, and process your API input and output." (Gemini API Additional Terms of Service)
"Do not submit sensitive, confidential, or personal information to the Unpaid Services." (Gemini API Additional Terms of Service)
So customer emails, intake forms, invoices or patient and client notes belong on a paid key only. On the paid tier, Google's pricing page marks content as "not used to improve our products".
How do you stretch a free quota?
Use it for what it is good at: testing prompts before you build anything.
- Check your real limits first. Open the rate limit page in AI Studio and note RPM and RPD for the model you use.
- Send fewer, bigger requests. One call that writes ten product descriptions uses one request, not ten. Stay under the TPM cap.
- Save answers you reuse. Store a reply (a FAQ answer, a service blurb) instead of asking the model again.
- Pick the smallest model that does the job. Flash-Lite handles sorting, tagging and short rewrites; save bigger models for drafting.
- Run heavy jobs after midnight Pacific. That is when the daily count resets.
- Plan for the error. When a limit is hit, the call fails. Any tool that runs on its own needs to wait and retry, or switch to a backup model, ideally from a second AI provider, rather than silently skip work.
When should you pay?
Pay when any of these is true: customers see the output, the task runs every day on a schedule, you send personal or client data, or you need a Pro-class, image or live-search model. Paying is cheap at small-business volume, but the setup has rules worth knowing (Gemini API docs: Billing):
| Step | What Google requires |
|---|---|
| Upgrade | Link a billing account in AI Studio and prepay at least $5 in credits |
| Credits | Expire after 12 months and are non-refundable |
| Tier 1 | Starts once billing is linked; monthly spend cap $250 |
| Tier 2 | After $100 paid and 3 days from first payment; cap $2,000 |
| Tier 3 | After $1,000 paid and 30 days from first payment |
| Balance hits $0 | All API keys on that billing account stop until you add credits |
Two price changes are already scheduled. Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026, then doubles to $1.50 and $7.50 on January 1, 2027. Gemini 4 Carbon doubles on December 1, 2026 (Gemini API docs: Pricing). On a paid key, the Batch API runs jobs that can wait at half the standard rate.
Set a project spend cap on the Spend page in AI Studio. Google calls it experimental and says you can still overspend for about 10 minutes, so set it below what you can actually afford.
Publishing an app you built in AI Studio is separate: eligible accounts can put up to 2 full-stack apps on Google Cloud's Starter Tier without a billing account (Gemini API docs: Billing).
Is it available in Canada?
Yes. Google lists both the United States and Canada among the countries where the Gemini API and AI Studio work (Gemini API docs: Available regions). Prices are listed in US dollars.
Example
A landscaper drafting blog posts
The owner tests prompts in AI Studio for free, then moves to a paid key. Thirty posts a month at about 5,000 input and 3,000 output tokens each on Gemini 3.8 Flash costs about $0.45 a month, or about $0.90 after the January 2027 increase. The $5 prepay covers months of use.
A dental office answering intake emails
Patient emails are personal data, so the free tier is ruled out from day one, whatever the volume. The office runs the tool on a paid key with a $20 project spend cap.
A retailer with a nightly product-tagging job
The job tagged 2,000 products a night on the free tier until its daily cap dropped and half the products went untagged. Grouping 20 products per request cut calls to 100 a night; a paid Flash-Lite key with a wait-and-retry rule removed the risk.
FAQ
Does making a new API key reset my limits?
No. Limits belong to the project, not the key. A new key in the same project shares the same quota.
Why do I get a rate limit error when I'm under my daily cap?
You probably hit the per-minute limit on requests or tokens. Space calls out or send fewer, larger requests.
Is chatting in Google AI Studio free?
Yes, direct use of AI Studio is free, but it counts as unpaid use under Google's terms, so the same human-review rule applies to what you type.
Will my free-tier limits go back up?
Google promises nothing. Its docs say limits can change with your tier and account status, and staff have said they adjust free limits based on demand.
Do I need a developer to switch to paid?
Not for billing: linking a billing account and buying credits is done in AI Studio. Changing which model a tool calls, or adding retry logic, usually needs whoever built the tool.
Different for you?
- Deciding which marketing tasks can run without a person watching: AI marketing automation.
- Splitting work between AI and your team: what to automate and what humans own.
- Using Gemini output on your website or blog: keeping AI writing accurate and on-brand.
- Wondering whether an AI agent beats a script that calls an API: agentic AI explained.
- Want leads followed up without building it yourself: see conversion.