Gemini API Rate Limits: RPM, TPM, RPD & Usage Tiers

Gemini API rate limits do more than cap request volume. They decide whether a traffic spike stays inside the minute-level quota, whether token-heavy prompts can run at scale, and when a project will start returning 429 RESOURCE_EXHAUSTED errors even though daily capacity is still available. T

This article brings those quotas into one reference. It lists the current Gemini API RPM, input TPM, and RPD limits for the main text models, then breaks out the separate ceilings for image generation, speech, embeddings, Live API, Batch jobs, and related workloads so you can compare tiers and plan around the limits that actually apply to your project.

How Gemini API Usage Tiers Work

The usage tier belongs to the API project and is determined through its linked Cloud Billing account. Tier 2 and Tier 3 qualifications use cumulative Google Cloud payments and the elapsed time since the first successful payment.

Gemini API Rate Limits by Model and Tier

RPM = requests per minute. Input TPM = input tokens per minute. RPD = requests per day. K = 1,000. M = 1 million. B = 1 billion. 

Free Tier

Gemini 4 Argon and Gemini 3.1 Pro require paid API access.

Tier 1

Tier 2

Tier 3

API Model IDs

Specialized Gemini API Rate Limits

Image Generation

Text-to-Speech and Transcription

Embeddings

Omni Models

Live API

Video and Music Generation

Batch API: Enqueued Tokens

Priority Inference and Grounding

Priority inference has a default rate limit of 30% of the standard limit for each model and tier. Its requests also count toward the normal interactive quota. For paid Gemini API usage with Gemini 3 and newer models, Grounding with Google Search includes 5,000 free search requests per month, shared across eligible models, followed by $14 per 1,000 search queries. One API request can trigger multiple billable searches. Google Maps grounding has its own 5,000-request monthly free allowance and the same $14-per-1,000-searches price.

Spend-Based Rate Limits

Spend-based limits use a rolling 10-minute window, distinct from RPM and input TPM. Their application depends on billing history and account standing.

How RPM, Input TPM, and RPD Are Counted

Gemini API model limits are per project. Multiple API keys in one project use the project’s quota. RPD resets at midnight Pacific time. A model’s context window describes capacity within an individual interaction, while input TPM measures traffic across requests.

How to Fix 429 RESOURCE_EXHAUSTED Errors

A 429 RESOURCE_EXHAUSTED response indicates a rate limit or spending constraint. Identify the affected model and quota: brief RPM bursts can be retried after the minute window, while exhausted RPD allowances require the daily reset or a higher eligible tier.

Python SDK Retry Behavior

The Google Gen AI Python SDK does not retry HTTP requests by default. Configure retry options to retry eligible 429 and transient server errors with increasing delays.

from google import genai
from google.genai import types
client = genai.Client(
    http_options=types.HttpOptions(
        retry_options=types.HttpRetryOptions(attempts=5)
    )
)

Frequently Asked Questions

Does creating another API key increase Gemini rate limits?

No. RPM, TPM, and RPD limits are applied to the project. Keys in the same project share its limits.

When does a daily Gemini API request limit reset?

RPD resets at midnight Pacific time.

How soon do higher API tiers take effect?

After billing activation, Free-to-Tier-1 upgrades normally take effect immediately. Eligible subsequent tier changes generally take effect within 10 minutes.

Does a Google AI Pro subscription raise Gemini API quotas?

No. Google AI Pro and Ultra provide higher allowances in the AI Studio web UI, not for direct API-key requests. Gemini API project tiers use Cloud Billing.

Alternatives & Related Resources

• Gemini API Free Tier Limits: Requests Per Day & Setup
• OpenAI API Rate Limits: RPM, TPM, Tiers, and 429 Errors
• AI LLM API Pricing: GPT, Gemini, Claude & More
• LLM Context Window Comparison: GPT, Claude, Gemini & More