The Gemini API Free Tier allows 20 requests per day for Gemini 3.8 Flash and Gemini 3.6 Flash, and 500 requests per day for Gemini 3.5 Flash-Lite and Gemini 3.1 Flash-Lite. Standard API calls are free, including input and output tokens, within those rate limits.
The daily allowance is only one part of the quota. Google also limits how many requests and input tokens you can send each minute. Flash-Lite gives a small Python script far more daily calls for classification, extraction, and repeated prompt testing, while Gemini 3.8 Flash is available for more demanding coding and multimodal tasks.
Gemini API Free Tier daily limits
These are the Free Tier limits for standard Gemini API requests. RPM means requests per minute, TPM means input tokens per minute, and RPD means requests per day.
| Gemini model | RPM | TPM | RPD |
|---|---|---|---|
| Gemini 3.8 Flash | 5 | 250,000 | 20 |
| Gemini 3.6 Flash | 5 | 250,000 | 20 |
| Gemini 3.5 Flash-Lite | 10 | 250,000 | 500 |
| Gemini 3.1 Flash-Lite | 10 | 250,000 | 500 |
How the free quotas are counted
The quotas apply per project and model. Creating another API key in the same Google Cloud project does not double the allowance. Reaching RPM, TPM, or RPD can trigger a 429 RESOURCE_EXHAUSTED error, even when the other two limits have capacity. Daily RPD resets at midnight Pacific time.
The figures above come from the Free Tier comparison in Google AI Studio Rate Limits. If your project belongs to a different usage tier, its active quotas appear there. The table gives the free allowances directly, and the link is only needed to inspect your own project.
Which Gemini models have changed?
On October 8, 2026, Google deprecated gemini-3.7-flash and gemini-3.5-flash. Calls to those old model IDs now run on gemini-3.8-flash and gemini-3.6-flash, respectively. New integrations should use the replacement model IDs.
Gemini 2.5 Flash, Flash-Lite, and Pro have a different restriction. They haven’t been retired, but API access is limited to developers who previously used them. New projects should use current Gemini 3.x models.
What is included in the Gemini API Free Tier?
A Free Tier project can make standard text-generation requests without linking a Cloud Billing account. The standard input and output tokens of Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, and 3.1 Flash-Lite cost $0 on the Free Tier. Your requests must stay within the RPM, TPM, and RPD allowances shown above.
Google AI Studio is available for creating an API key and testing prompts. Standard Free Tier access does not include every premium feature: for these current Flash and Flash-Lite models, Google Search and Maps grounding are not part of unpaid standard API usage. Data-use conditions also differ between free and paid API projects.
Gemini API Free Tier vs. Paid Tier limitations
Paid access raises rate limits and unlocks models such as Gemini 4 Argon. For a concrete comparison, the Tier 1 limits displayed in AI Studio are 10,000 RPD for Gemini 3.8 Flash and 150,000 RPD for Gemini 3.5 Flash-Lite. The corresponding Free Tier limits are 20 and 500 RPD. Tier 1 requires a linked billing account, and standard API tokens are billed at the model’s paid rates.
| Free Tier | Paid Tier | |
|---|---|---|
| Daily Flash allowance | 20 RPD (3.8 / 3.6) | Up to 10,000 RPD at Tier 1 |
| Daily Flash-Lite allowance | 500 RPD (3.5 / 3.1) | Up to 150,000 RPD at Tier 1 |
| Standard tokens | $0 for eligible models | Charged by model |
| Gemini 4 Argon | Unavailable | Available |
| Cloud Billing | Not required | Required |
| Google Search grounding | Unavailable on these models | Paid terms and quotas |
| Data use | Unpaid-service terms | Not used to improve Google products |
Is Gemini 4 Argon available on the Free Tier?
No. Gemini 4 Argon (gemini-4-argon) is a paid-only Gemini API model. A Google AI Pro or Ultra subscription in the consumer Gemini app does not give your API project free Argon requests.
Which free Gemini model should you use?
Gemini 3.5 Flash-Lite has 500 free daily requests and a 10 RPM allowance. It works well for repeating short prompts, categorizing support messages, extracting fields from text, and translation. Its API model ID is gemini-3.5-flash-lite.
Gemini 3.8 Flash has 20 free daily requests and a 5 RPM allowance. It is better suited to more demanding coding, document analysis, and multimodal prompts. Use gemini-3.8-flash in the API. Gemini 3.6 Flash and Gemini 3.1 Flash-Lite remain available if you need those specific models.
A larger context window does not increase RPM, TPM, or RPD. The LLM Context Window Comparison: GPT, Claude, Gemini & More has the model input and output limits for longer documents or codebases.
How to use the Gemini API for free
The example below uses Gemini 3.5 Flash-Lite to classify three fictional support messages. It needs a Google AI Studio API key, Python, and the current google-genai package.
1. Create your Gemini API key
Open Google AI Studio API keys, select or create a project, and generate an API key. You can make eligible Free Tier requests without linking a billing account.
Store the key outside your Python source file. Do not expose API keys in a public repository or browser-side JavaScript.
2. Install the Google Gen AI SDK
pip install -U google-genai3. Set your API key
In Windows PowerShell:
$env:GEMINI_API_KEY="YOUR_API_KEY"On macOS or Linux:
export GEMINI_API_KEY="YOUR_API_KEY"4. Send your first request
Save the following code as gemini_test.py:
from google import genai
client = genai.Client()
prompt = """
Classify each fictional website support message as
Billing, Bug, or Feature Request.
1. My invoice shows two charges for this month.
2. The upload button stops responding after I select a PDF.
3. Please add an option to export reports as CSV.
"""
response = client.models.generate_content(
model="gemini-3.5-flash-lite",
contents=prompt,
)
print(response.text)5. Run the script
Use the terminal where you set your API key:
python gemini_test.pyOn systems where Python is started with python3, run python3 gemini_test.py. A successful call returns three message classifications in your terminal.
How RPM, TPM, and RPD affect your free requests
Gemini checks all three quotas for every standard API call. Under Gemini 3.5 Flash-Lite, for example, the Free Tier allows 10 requests per minute, up to 250,000 input tokens per minute, and 500 requests per day. A burst of 11 requests in one minute exceeds RPM even though hundreds of daily requests are available.
Long documents, transcripts, and concurrent calls consume TPM. RPD counts how many requests you make during the day, regardless of how short the prompts are. Gemini 3.8 Flash accepts up to 1,048,576 input tokens per request, yet its Free Tier TPM limit is 250,000. A long request can exceed TPM even if the prompt fits within the model’s context window.
What to do when Gemini returns a 429 error
A 429 RESOURCE_EXHAUSTED error indicates that a rate or other applicable usage limit has been reached. The correct fix depends on which limit the request exhausted:
| Limit reached | What to do |
|---|---|
| RPM | Reduce the request frequency and retry after the minute window recovers |
| TPM | Use smaller prompts or reduce concurrent large requests |
| RPD | Wait until midnight Pacific or use a suitable model with its own available quota |
| Repeated 429 | Check whether the project is on Free Tier or a paid tier |
| 503 service error | Retry after a delay when the service is temporarily unavailable |
Exponential backoff works for temporary rate pressure. It cannot restore a fully used daily RPD allowance before the reset.
Privacy and data use on the Gemini API Free Tier
For unpaid Gemini API requests outside the European Economic Area, Switzerland, and the United Kingdom, Google can use prompts and responses to improve its products, and human reviewers can inspect some submissions. For free API tests, avoid private customer messages, confidential documents, credentials, and other sensitive material.
Paid Gemini API services do not use your prompts and responses to improve Google products. For developers located in the European Economic Area, Switzerland, or the United Kingdom, the paid-services data-use provisions also apply to unpaid requests. A Gemini API client offered to people in those regions must use Paid Services. These conditions are set out in the Gemini API Additional Terms of Service.
FAQs
Are the 20 and 500 free daily requests shared across models?
The rate-limit figures are model-specific and applied to your project. Gemini 3.8 Flash has a 20 RPD Free Tier allowance, and Gemini 3.5 Flash-Lite has a 500 RPD allowance. Creating several API keys within one project does not multiply either model’s quota.
Can I use a Google AI Pro subscription to get a larger Gemini API Free Tier?
No. Google AI Pro and Ultra subscriptions for the Gemini app are not Gemini API billing tiers. API rate limits depend on the Google Cloud project’s Free, Tier 1, Tier 2, or Tier 3 status.
Are Gemini image-generation and Live API limits included in this table?
No. The table applies to the standard Gemini text-output models listed. Image-generation and Live API models have their own rates, eligibility, and units. Do not use the 20 or 500 RPD figures to estimate image or real-time audio usage.
Related Resources
- Google AI Timeline: Gemini 4 Argon, Models & Product Releases
- OpenAI API Rate Limits: RPM, TPM, Tiers, and 429 Errors
