Google AI Studio free tier: API limits, billing, and upgrade checks
Google AI Studio free tier limits vary by model and project. This guide explains RPM, TPM, RPD, quota checks, 429 handling, data-use terms, and the point at which linking billing changes your risk.
Google AI Studio free tier is useful for prototypes and low-volume scripts, but it is not one fixed monthly bucket shared by every model. Google documents Gemini API limits across requests per minute, tokens per minute, and requests per day, with values that vary by model, project, and usage tier. Google AI Studio billing depends on the project and key you connect. Before you ship a small application, check the quota shown for the actual project in AI Studio.
Separate the web playground from the API tier
AI Studio is the development surface. The Gemini API is the interface your application calls. Google’s billing guide says new accounts start on the Free Tier, which gives access to certain models up to those models’ free-tier rate limits. That is different from message limits in a consumer Gemini app. Successful chats in the web UI do not prove that an API workload has the same allowance.
For a small support prototype, create a separate Free Tier project and keep test and production keys apart. The documented rate limits apply per project, not per API key, so adding keys to one project does not create a second project quota. Preview and experimental models can have tighter limits. Record the exact model name, then recheck the Gemini API pricing page and the rate-limits guide before release.
What RPM, TPM, and RPD actually block
RPM is requests per minute. TPM is input tokens per minute. RPD is requests per day. A workload can touch more than one dimension. Many short prompts can hit RPM first, while long context or long outputs can hit TPM. A low request count does not guarantee that token usage is below the limit.
Google says daily request quotas reset at midnight Pacific time. Teams troubleshooting from Singapore, Beijing, or Europe should convert log timestamps before deciding that a quota has reset. When the API returns 429, capture the model, project, timestamp, and response details. Reduce concurrency or context size, use bounded backoff, and stop when the retry budget is exhausted.
Google also documents spend-based rate limits for higher usage tiers. Free is shown as N/A, while Tier 1, Tier 2, and Tier 3 have different ten-minute caps. Those caps protect billing. They are not a universal daily request allowance. For model-specific quotas, read the active quota panel in AI Studio.
Treat data use as a separate decision
The Gemini API pricing page labels Free and Paid tiers differently for data use. Free content may be used to improve Google products, while the Paid listing says content is not used for product improvement. That difference matters when prompts contain source code, customer conversations, or unreleased product details.
