Gemini API rate limits: 4 tiers, spend caps and reset time

In brief

Gemini API rate limits run on three counters, requests per minute, input tokens per minute and requests per day, set per project by tier, as of November 2026. Tier 1 needs billing, Tier 2 $100 paid plus 3 days, Tier 3 $1,000 plus 30 days. Daily limits reset at midnight Pacific. Google prints no per-model request limits for text models.

What are the Gemini API rate limits, tier by tier?

Gemini API limits depend on your tier, and Google sets the tier by how much you have paid. Here is what Google's rate limits page says, as we read it on 14 November 2026. The Free tier is for any active project, and some projects need verification first. The three paid tiers upgrade you on their own once you meet the rule.

A chrome Gemini star hovering over four glass funnels of rising width on a dark studio floor, each one pouring a thin stream of light into a terracotta tray.
Tier, as of November 2026How you reach itBilling account capSpend limit per 10 minutes
FreeAn active project or free trialNoneNot applied
Tier 1Set up and link an active billing account$250$10
Tier 2$100 paid in total, plus 3 days from the first successful payment$2,000$200
Tier 3$1,000 paid in total, plus 30 days from the first successful payment$20,000 to $100,000 or more$200
The Gemini API billing cap jumps 8 times from Tier 1 to Tier 2
$250$2,000$20,000Tier 1Tier 2Tier 3, lowest figure

Google counts the $100 and $1,000 as total spending on Google Cloud services for the linked billing account, not on the Gemini API alone. Some search summaries give Tier 2 a $50 spend limit per 10 minutes. Google's table says $200, as read on 14 November 2026. Google also says the spend limits may not apply to every account, depending on billing history and standing. Treat the AI Studio page as the last word for your project. Google says specified limits are not guaranteed, and that actual capacity may vary, so a number that held last week can shift without any change on your side. Free to Tier 1 is usually instant. Later upgrades land within 10 minutes of meeting the rule. The Projects page in AI Studio shows which tier each project is on. Google adds that a tier can occasionally be denied because of other review factors, and that its limits aren't guaranteed, since capacity may vary.

What do the three limit types count?

Google counts three things: requests per minute (RPM), input tokens per minute (TPM) and requests per day (RPD). Pass any one of them and you get an error, even when the other two are fine. Google's own example is a project with an RPM limit of 20. Its 21st request inside a minute fails, no matter how few tokens it carries.

Limits apply to a project, not to an API key. Two keys in the same project share one pool, so a second key won't double your room. Limits are counted for the project, so that is the unit to plan around. Some limits only exist for some models. Image models add images per minute (IPM), a few models carry a tokens per day cap, and built-in tools have their own. Grounding with Google Search on Gemini 3 and newer has a default of 1,500 requests a day per project.

Preview and experimental models get tighter limits. Google says so in plain words: "Rate limits are more restricted for experimental and preview models." Gemini 3.1 Pro is still a Preview model, so plan for less room there than the tier table suggests. Gemini API pricing lists which models are which.

What are the Gemini API rate limits per model?

Google doesn't publish per-model RPM, TPM or RPD numbers for its text models in its docs. As of 14 November 2026, the page sends you to your live numbers in AI Studio: "For details on those rate limits, see the AI Studio Rate Limit page." Any blog that quotes you a fixed table is repeating a snapshot, and Google says plainly that a published limit is "not guaranteed".

Two models do get exact figures on the page. Both come with a Free column, which makes them the only rows where you can compare tiers side by side.

Model, as of November 2026FreeTier 1Tier 2Tier 3
Gemini Omni Flash, requestsNot supported8 a minute, 40 a day16 a minute, 200 a day40 a minute, 2,000 a day
Gemini Decision 1, requests200 a minute, 10,000 a day1,500 a minute, 50,000 a day5,000 a minute, no daily cap15,000 a minute, no daily cap
Gemini Decision 1, input tokens50,000 a minute1,000,000 a minute4,000,000 a minute10,000,000 a minute

The Batch API has the most numbers of all. Every batch job set shares 100 concurrent requests, a 2 GB input file and 20 GB of file storage, and each model gets a ceiling on enqueued tokens. Gemini 3.8 Flash can hold 3,000,000 tokens in the queue on Tier 1, 400,000,000 on Tier 2 and 1,000,000,000 on Tier 3. Gemini 4 Argon gets 5,000,000, 500,000,000 and 1,000,000,000. Priority traffic runs at 0.3 times the standard limit for each model and tier by default, and it also counts toward your overall limit.

What is the Gemini API free tier rate limit?

The free tier has real limits, but Google doesn't print them for text models, so the only honest answer is to open AI Studio. Sign in, open the rate limit page, and the dashboard shows the RPM, TPM and RPD for each model your project can call. It updates as your tier changes.

That means the question is less about the number and more about what the free tier is for. Google's billing docs say "New accounts begin on the Free Tier," and the free tier's content "used to improve our products", per the pricing page. It suits a prototype. It won't carry a product with paying users. Is the Gemini API free goes through which models have a free tier at all, and Gemini free vs paid covers the app instead.

On the free tier there's also no spend limit per 10 minutes, since you haven't put money on the line. The first thing you'll hit is a request counter. Moving to Tier 1 only takes a linked billing account, and Google says that step is usually instant.

When do Gemini API rate limits reset?

Daily limits reset at midnight Pacific time. Per-minute limits come back as the minute rolls over. The spend limit works on a rolling 10 minutes, so it recovers as older spend leaves the window. Billing caps are different. They reset on the 1st of the month, and Google says that when a cap is hit it pauses "service for all projects linked to that billing account" until then.

So a 429 at 11 pm Pacific isn't the same as a 429 at 1 am. Time matters. If you run out of requests per day late in the evening, the fix may be to wait a few hours, because the daily counter clears at midnight Pacific while a per-minute counter clears within the minute and the 10 minute spend window clears as older spend ages out of it. The check for it is the quota graph in AI Studio. If the daily bar is full, wait for midnight. If it isn't, you hit a per-minute limit and a short pause will do.

How do you fix a Gemini API 429 error or raise the rate limit?

A 429 means a limit ran out, and the fixes below are in the order that works most often. Google's docs name the spend-based case: "If you hit a spend-based rate limit, the API returns a 429 RESOURCE_EXHAUSTED error." Its retry guidance covers the rest, and it recommends exponential backoff with random jitter, so that clients don't all retry in the same second.

  1. Wait and retry with backoff. Retry only 429, 408 and 5xx, and set a maximum number of tries. The official Python SDK already retries transient errors up to four times.
  2. Check which limit you hit. Open the AI Studio rate limit page. A full daily bar means wait for midnight Pacific. A full per-minute bar means slow down.
  3. Shrink the request. Smaller context windows and shorter outputs lower the tokens per minute and the spend per 10 minutes.
  4. Move up a tier. Link billing for Tier 1. Pay $100 and wait 3 days for Tier 2. The limits rise on their own, usually within 10 minutes.
  5. Move the work to Batch. Jobs that can wait use their own queue, not your live limits.
  6. Ask Google for an increase. Paid users can use the request form linked on the rate limit page. Google guarantees nothing.

If you reach the wall as a person in the Gemini app instead of through the API, the limits are different ones. Gemini too many requests and Gemini usage limit reached cover those.

What if the app is the problem and not the API?

If you only want to chat with Gemini, not build on it, you don't need an API key or a tier at all. Whizi's free plan gives you 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, 2 pictures a month and 5 photo uploads a day, with no card. Pro carries Claude, GPT and Gemini, and its first week is $1.99.

It won't replace the API. Whizi doesn't give you a key, and it doesn't run your code against Google's limits. It fits the reader who hit a wall while using a model to think or write, not the one running a pipeline. If your job is calling Gemini from software, stay on this page and raise your tier.

Workflow checklist
  • Open the AI Studio rate limit page and read the live RPM, TPM and RPD for your model, since Google doesn't print them in the docs
  • Retry a 429 with exponential backoff and jitter, with a maximum number of tries
  • Check the Projects page to see which tier the project is on before you plan around a number
  • Wait for midnight Pacific if the daily request bar is full
  • Move jobs that can wait to the Batch API, which has its own queue limits
Common questions

Frequently asked questions

What are the Gemini API rate limits?

Three counters set per project and per tier: requests per minute, input tokens per minute and requests per day, as of November 2026. Google prints exact numbers for only two models, Gemini Omni Flash and Gemini Decision 1. For text models it sends you to the AI Studio rate limit page for your live numbers. Gemini API pricing covers what each model costs.

How do I get a higher Gemini API rate limit?

Move up a tier, which happens on its own. Tier 1 needs a linked billing account. Tier 2 needs $100 paid and 3 days from the first payment. Tier 3 needs $1,000 paid and 30 days. Paid users can also send Google a request form, though Google says no increase is guaranteed.

When do Gemini API rate limits reset?

Requests per day reset at midnight Pacific time. Per-minute limits recover as the minute passes, and the spend limit works over a rolling 10 minutes. A billing account that hits its monthly cap stays paused until the 1st of the next month.

Are Gemini API rate limits per key or per project?

Per project, not per API key. Google says so on its rate limits page. Creating a second key inside the same project gives you no extra room, because both keys draw from the same pool.