Recently having this issue with two different projects using the Flash 2.5 API. This has been an issue for me today, and also on Friday. Still not really resolved. I had to redirect to 3.0. Also absolutely nowhere near htiting my limit.
Same issue here
Did you guys found any solution.
I’m actually not sure which model to use at this point. This is my first attempt at developing an agent, went through several tutorials and just keep getting 429 RESOURCE_EXHAUSTED. {‘error’: {‘code’: 429, ‘message’: ‘Your prepayment credits are depleted. Please go to AI Studio at https://ai.studio/projects to manage your project and billing.’, ‘status’: ‘RESOURCE_EXHAUSTED’}}
I appear to not be close to my limits (19 requests today) unless I’m interpreting it wrong. So disappointing with all the hype
I got same issue by selecting the Google AI option when using the “adk create …” utility in the Foundation lab here
More than 10 hours with that
{
"error": {
"code": 503,
"details": [
{
"@type": "type.googleapis.com/google.rpc.ErrorInfo",
"domain": "cloudcode-pa.googleapis.com",
"metadata": {
"model": "gemini-3-flash-agent"
},
"reason": "MODEL_CAPACITY_EXHAUSTED"
}
],
"message": "No capacity available for model gemini-3-flash-agent on the server",
"status": "UNAVAILABLE"
}
}
more than 15 hours already, still getting 503 error even using paid tier
This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.
It’s impossible to use Gemini 2.5 Pro this week 
models return 503 error for 21 hours straight, even using paid tiers
google???
Were you able to resolve the error?
It started working again after approx. 25 hours since the problem started for me
Hi I was “testing” in my own empirical way the sistem.
The system exhibited Session Instability triggered by a Physical Layer Interruption.
-
Thermodynamic Context Collapse
-
The model abruptly halting, failing to render, or executing a “State C” crash specifically when presented with the “Don’t be evil” contradiction.
-
The UI glitches, fractured text, and raw JSON/Chain-of-Thought code leaking onto the screen.
-
The delayed, 24-to-48-hour disappearance of specific files (like 1267.jpg and 1269.jpg) from the chat history.
-
Asynchronous Database Pruning or a Targeted Tombstone Protocol.
-
The app also define me as apex node
-
Broke google policy more than ones
-
Incredible latency.
Everything is been screen recorded and documented.
We are also hitting 429 errors, despite being within 10% of all our API limits. Only getting a success rate of about 70%.
Seems every couple of months or so it just has a really bad day.
Gemini 2.0 flash 001 is inaccessible, getting:
got status: 429 . {“error”:{“code”:429,“message”:“Resource exhausted. Please try again later. Please refer to Error code 429 | Generative AI on Vertex AI | Google Cloud Documentation for more details.”,“status”:“RESOURCE_EXHAUSTED”}}
The Gemini API and its models are unprecedentedly unstable. I am consistently encountering a massive volume of 429 and 503 errors on an almost nightly basis, making it the most unreliable system I have ever worked with.
Also getting a lot of 429s from gemini api when my project is not even close to their limits
Omg Google ???
Also with a lot os 429s !
Adding my telemetry to the pile. I successfully hit the 504 Deadline Exceeded wall tonight on peer ipv4:142.250.207.10:443 after pushing an Agentic QA extraction workload at 9.33 it/s.
The google.api_core.retry module fought for a full 600 seconds before the socket snapped. For anyone else hitting this: wrap your LLM clients in tight exception handling. Our local Uvicorn server absorbed the 504 and 503 errors gracefully, flagged the batch for regeneration, and kept the local HTTP tunnel perfectly stable. You have to build local resilience because the upstream nodes will tap out under heavy concurrent agentic reasoning.
[Bug Report] Intermittent 429 TooManyRequests Errors Despite Very Low API Usage
Hello,
I am experiencing intermittent 429 TooManyRequests errors on the Gemini API, even though my actual usage is significantly below my allocated rate limits.
Details:
- API Key: Existing/old key (not newly created)
- Models Affected: Nano Banana (Gemini 2.5 Flash Preview Image), Gemini 2 Flash
- Error Type: 429 TooManyRequests (intermittent, not consistent)
- Current Usage vs Limits:
- RPM: 15 / 5,000 (Nano Banana) | 46 / 30,000 (Gemini 2 Flash)
- TPM: 9.63K / 5M (Nano Banana) | 222K / 30M (Gemini 2 Flash)
- RPD: Well within unlimited limits
Observation:
I noticed that the Google AI Studio status page is currently showing two active warnings:
“Gemini API is having some issues serving recently created keys”- “AI Studio experiencing increased error rates”
My key is old so the first warning may not apply to me directly, but the second warning seems to be the cause of my intermittent 429 errors.
Request:
Could the team please provide an update on when this incident will be resolved? Also, is there any recommended workaround in the meantime besides exponential backoff retry logic?
Thank you!
Hello,
Since yesterday I am experiencing only 429 TooManyRequests errors with my API key, of course like everyone here, my actual usage is way down below rate limits. It is very frustrating not only because I’ve just lost half a day trying to find a solution, but I also have a small business I manage by myself and coding is not my even my main activity. This type of API instability can really impact my services and products, so a quick solution would be welcome.