The Gemini API global status panel is reporting severe incidents of ‘increased error rates’ and ‘issues serving recently created keys’. This is generating a flood of 429 and false 404 errors in my project, even though I have quota.
Omg! You tried contact with the support ?
they didn’t help here!
So we have to wait ? thats so frustrating!!!
yeah really frustrating tbh. i contacted support already, after some back and forth they escalated it to the internal team. still waiting for any update though…
i’ve move to vertex API to see ifs go better! 
did it actually work better on vertex?
no! it’s the same 
Same here, looks liks its problem with flash 2.0
yes, I analyzed the logs! I’ve changed for 2.5-flash-lite and it’s working!
I have met the same problem.
All of my service depend on 2.0 flash.
It’s too crazy
Hello everyone,
we are experiencing a lot of issues on Gemini API endpoints while processing images, anyone is experiencing same issues or fixed yet?
Thanks
The “internal team” wasnt able to solve anything for me. How about you?
Adding data points: I’m on Paid 2 / Tier 2-eligible (well past $100 cumulative spend, all qualification criteria met), but my AI Studio Rate Limit page still shows Tier 1 with “spending cap exceeded” 429s on Gemini 3.1 Flash Image (Nano Banana 2). All 4 of my API keys on the same billing account return identical errors within ~200ms. I spun up a fresh Google account with new keys — those returned different errors (“prepayment credits depleted”) despite never being funded. Google Cloud Billing Support just closed my case saying Gemini API tier issues aren’t in their scope. This all lines up with a serving-layer / infrastructure incident rather than genuine billing or quota enforcement — anyone else seeing the “spending cap exceeded” variant specifically?
Hate to say it but until the model goes GA or they roll out 4.0 one day we may never see a solution to this. I’m Tier 3 and can only process 3-4 images per minute using 3.1 Preview without breaking it during peak times. And sometimes even during off peak times it breaks. Lots of good error handling and patience is all you can do. Hoping they’re building up those data centers to handle actual commercial production rates.
Yes, I’ve changed for 2.5 flash lite! But I wanna move for Gemini too, you guys are going to use what IA ?
were you able to fix this? I am getting 503 error on all nano banan models
I am on a free level at no cost, but I make sporadic and very occasional use, all text.
With Flash version 2.5, Flash has been returning more and more 503 errors since approximately April 13 and is becoming unusable.
The status shown by Google AI Studio is curious and embarrassing.
The fact of having a free level does not exempt us from providing good service, and I am not exceeding requests.
I think that in the end it will die of success because the demand exceeds the service capacity and they will solve it by restricting tokens and increasing the subscription cost.
Is it not possible to segment by areas or locations or develop a queue of requests?
It’s frustrating to have 503 errors over and over again.
I’m seriously thinking about setting up a Local AI system, it probably won’t be as powerful, but I think it will pay off in the long run.
They will sell us that the XX.y model is the most capable and best, but honestly… Do we need a Ferrari to go to the corner store (which stalls the engine with 503 more and more often)?
It will depend on the real use cases, but right now Google AI Studio is a shame.
Regards
