Gemini API Rate Limits: RPM, TPM, RPD, Tiers, and 429 Errors

Gemini API limits are defined across multiple dimensions and can vary by model, usage tier, project, and account status. A copied RPM table can become stale; Google AI Studio is the best place to view the active limits that apply to your project.

Facts checked August 10, 2026. Limits and plan names change frequently. The official account page linked in each guide is the source of truth for the limits attached to your account.

Gemini API Rate Limits: RPM, TPM, and RPD

• RPM: requests per minute.
• TPM: tokens per minute.
• RPD: requests per day where a daily quota applies.

A request succeeds only when it fits every applicable limit. Low request volume can still hit TPM with large prompts, while tiny requests can exhaust RPM first.

Where to Find Your Actual Gemini Limits

1. Open Google AI Studio.
2. Select the project and API key used by the application.
3. Open the Rate Limit view to see model- and tier-specific limits.
4. Open Usage to compare recent activity with those limits.

Rate limits are generally enforced per project, not separately for every API key. Creating another key in the same project therefore does not create a new independent quota pool.

Why Gemini Returns HTTP 429

A 429 usually indicates that one of the active request, token, or daily quotas was exceeded. Check that the application is using the expected project, account for usage reporting delay, and inspect every quota dimension. If the active dashboard shows remaining quota, retry with exponential backoff and investigate whether capacity or another service condition is involved.

Free vs Paid Gemini API Tiers

Paid tiers generally unlock higher limits as eligibility and billing requirements are met, but exact values depend on the model and current tier. Moving to paid access does not mean every quota disappears. Confirm the active values rather than assuming a paid project is unlimited.

How to Avoid Gemini Rate-Limit Errors

• Queue work and cap concurrency instead of sending an unbounded burst.
• Use exponential backoff with jitter for retryable 429 responses.
• Track tokens as well as requests.
• Batch compatible work when the API and workload support it.
• Request a tier increase only after measuring which dimension is actually constrained.

Gemini Rate Limits FAQ

Are Gemini limits per API key?

Google documents Gemini API quotas at the project level. Multiple keys in one project do not imply separate pools.

When does Gemini daily quota reset?

Use the active rate-limit view for the quota and schedule attached to your project rather than relying on a historical table.

Where is the official limit table?

Start with Google's Gemini API rate-limit documentation, then open the authenticated AI Studio Rate Limit page for current project values.