I am trying to use it to perform some processing of data, but it is highly unstable, i have retries and even with 3 retries it keeps failing, is there any instability? Also why are you guys deprecating it, if the costs of the next flash lite model are significantly higher than 2.5. Very disappointed.
I’m not even processing them in parellel they are running sequentially, this shouldn’t be happening
Same here is happening to me. I’m always below 40% success rate in the last days.
Same thing happening here, the last couple days it has gotten even worse
Hello all,
503 errors are due to our services being temporarily overloaded. Please see this pinned post for suggestions on how to handle this.
This answer is completely ridiculous, the models is not “temporarily overloaded”, how do you explain that yesterday and today i am still getting these errors, there very obviously is some problem here, this is completely unacceptable, we have implemented retries already, we attempt it 20 TIMES and it still manages to fail 20 times. And im sorry but refactor our whole flow just to use batch operations is a joke of an response, like yeah sorry we cant provide stability so please refactor everything you’ve done so that it runs in a more complicated and tricker way so that we can handle the load.
and when i check gemini status it says there is nothing wrong.
Yes, I’ve been seeing the same issue over last 3 days. Other models are working normally. It is like the resources allocated for gemini-2.5-flash-lite are not enough.
HI
Can you DM me your project ID, billing account ids ?
@Mustan_lokhand Happening also to us, we are seeing a lot of failures where even 3x retries do not help. We are seeing similar uptimes that Vercel AI Gateway is reporting for Gemini 2.5 Flash Lite (eg. around 40% success rate of requests). We are on tier 3 rate limits.
Not sure how AI studio status can be green because this is a service disruption.
Is it resolved for any of you? I am still getting lots of 503 errors
Also, Gemini 2.5 Pro getting a lot of failures.
Same with 2.5 pro Model
Can I also DM you my project ID? Is there any solution to this? Or can we get compensation? Some free credit?
Sometimes you get a 503 error message due to technical issues, and the Google team is the one that knows how to fix those error messages, like the 503 you’re getting when using Gemini 2.5 Flash Lite.
503 indicates high load, yet the uptime monitor shows no such issue.
Thanks all for reporting this. We escalated the issue and a fix was put in place. 503’s error rates should now be much lower.
