We are encountering “429 Resource Exhausted” errors when attempting to use Gemini 3 model, both through the API and AI Studio, despite remaining within our paid tier limits. Please investigate this issue and provide a solution to restore normal API and AI Studio functionality. All is well for the past few weeks but today we can no longer use the API normally.
Error: 429 RESOURCE_EXHAUSTED. {‘error’: {‘code’: 429, ‘message’: ‘Resource has been exhausted (e.g. check quota).’, ‘status’: ‘RESOURCE_EXHAUSTED’}}
The 429 Resource Exhausted error typically indicate that too many requests are being made in a short period. Even with a paid account, rate limits still apply, especially on a per-minute or per-region basis.
To address this:
Please check your per-minute and regional quotas in the Google Cloud Console under IAM & Admin > Quotas, filtering by generativelanguage.googleapis.com.
Ensure your billing account is correctly linked to the Gemini API project. Some users have found that quota enforcement can behave unexpectedly if the linkage isn’t fully established.
I need urgent help regarding the Gemini API. I am currently on the Paid Tier 1 plan, but I have been completely unable to use the API for the last 48 hours.
The Issue: Every request I make returns a 429 RESOURCE_EXHAUSTED error immediately.
The Contradiction: My Google Cloud/AI Studio usage dashboard shows “No Data Available” and 0 requests for the last hour/day. I clearly haven’t hit any rate limits because I haven’t been able to make a single successful request.
Details:
Error Message:{ "code": 429, "message": "Resource has been exhausted (e.g. check quota)", "status": "RESOURCE_EXHAUSTED" }
Plan: Paid Tier 1 (Pay-as-you-go)
Duration: This has been persisting for 2 days.
Impact: My productivity is at a standstill.
What I’ve checked:
I verified I am on the Paid Tier.
I verified my usage graphs are empty (screenshots attached).
I have waited and retried, but the block is permanent.
Has anyone else experienced a “phantom” quota limit on a paid plan?
I would just like an announcement that they are analyzing and looking into the problem. We can’t report it to anyone, we don’t know what’s happening or when it will be fixed - absolute disregard for the consumer. There’s no open support channel, there’s nothing, but it was my mistake to have built my product on top of a single LLM provider. I’m already going to rewrite everything to support other providers and remove Gemini.
I would suggest trying to authenticate via a Google Cloud account rather than the Gemini API Key if you’re encountering this problem and don’t know what else to try.
I believe the quota system on Google Cloud (Vertex AI) is different from the “AI Studio” system so I think this might help.
For those unaware, the background to this is that there’s two ways to access the Gemini models. The “AI Studio” (“Gemini API Key”) is an ad hoc system built specifically for Gemini. They wanted to have an easier on-ramp onto Gemini for developers who weren’t using the wider Google Cloud platform. However, this isn’t the only way to use Gemini. You can also use the Gemini functionality as part of Google Cloud Platform. The name of the Google Cloud Platform product group that has Gemini is “Vertex AI”. So instead of authenticating with a Gemini API Key, you can authenticate to Google Cloud Platform and use it that way.
The duplication is kind of crazy because they refuse to introduce any sort of sensible terminology for the distinction, and the different landing pages all pretend the other workflow doesn’t exist.[1] The non-Google-Cloud version of Gemini just calls itself “Gemini API”. But if you use Gemini within Google Cloud platform, obviously you are also using “the Gemini API”. And then there’s some functionality that only works if you authenticate one way or the other — but there’s no language to describe this difference, so the documentation just doesn’t talk about it. The documentation would have to say something like: “If you use the Gemini API in Vertex AI you can’t use client.files. To use that functionality you need to use the Gemini API.” I guess that’s embarrassing to write, so they just don’t write it.
It’s surprising to me that Google has invested billions into this company-critical venture, and this is how they manage the product.
We truly appreciate you flagging this issue and apologize for the inconvenience. Could you please provide the project number (not the project ID) via direct message?
Hi @Sonali_Kumari1 where do we send direct message? I cant see a message link on your page. Im also getting this
Timeout of 120.0s exceeded, last exception: 429 RESOURCE_EXHAUSTED. {‘error’: {‘code’: 429, ‘message’: ‘You exceeded your current quota, please check your plan and billing details. \n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 20, model: gemini-2.5-flash\nPlease retry in 45.06576809s.’, ‘status’: ‘RESOURCE_EXHAUSTED’, …
even when we’re using paid tier with the Gemini API key. The message also refers to the free tier not sure why when we’re on paid and we certainly passing the correct api key with the requests. We not making that many requests either probably 3-5 per hour and its been working fine till early this month.
You been posting this same message for many days on similar posts its making me think this is not something even being looked into at the moment? Its certainly not on an individual basis something is wrong in the server
It’s not just you. This is the issue that many are having, myself included. It’s not my code. And, it doesn’t affect me in the Gemini Studio. But, when I move to a Dev Server or Android Studio, this error pops up immediately.
same here, I am using GCP platform and I am in paid tier 2, I constantly get this error message when I am using gemini batch file api. I am also a user of GCP platform, the gemini api key is the one I created within my GCP project. I think the corresponding service name to query in GCP platform should be Generative Language API, when I use gemini batch file api, I am using gemini-2.5-flash model, from the gemini AI studio of from GCP dashboard, for this model, RPM,TPM are way below the quota, the batch file concurrent number is also not exceeding the 100 quota, the support replied me is: Enqueued tokens per model: This is the most likely quota you are hitting. There’s a maximum total number of tokens that can be in the queue for batch processing across all your active and pending batch jobs for a specific model (e.g., gemini-2.5-flash). Even if jobs are marked “completed,” they might take time to be fully cleared from the queue, still counting against this limit. Currently there is no option to monitor the “batch enqueued token count”.
You can create a feature request in the “AI Studio community”[1], so is this Enqueued tokens per model is the quota I am hitting the limit? Can you provide any way of query it usage?
I am also getting this error. We are on GCP using vertex ai. I have made maybe 10 requests total today and 3 of them were 429s. Does google expect us to ship production apps on Gemini with this kind of reliability?
I reckon the winds have calmed now, but an hour ago, the VertexAI was plagued with those 429 ‘resource exhausted’ errors on the Gemini 3 models.
Since it’s but a preview and not yet forged for the general folk, I suppose such instability is to be expected… but it’s a sorrowful thing nonetheless.
