Hello, I am writing to you because I use Gemini 2.5 Flash with a free plan, and I know that there is obviously a limit and that those who pay have priority, which is fair enough, but in the last 3-4 days it has been unusable, constantly overloaded, Since I already pay for a third-party service that uses Gemini to process articles with AI, I would now like to ask you to please resolve the reason for this excessive overload while maintaining the benefits for others who have paid plans. I am not asking for more than what I have; I just want everything to work as well as it has since I started using it last month. That’s all. Thank you for your attention, and I hope things get resolved soon
1 Like
Gemini was recently castrated again.
2 Likes
Among other things, by constantly regenerating the article, I use up more prompt attempts than I would if it worked on the first try, resulting in a rapid approach to the rate limit, allowing me to use the application for half or less of the time available. It’s really unbelievable. I hope there is a desire to improve things behind it, otherwise it would just be a disappointment.
I just had 11 503 errors with Flash in a row and that’s on a paid tier.
This is getting a bit ridiculous.
After a day when everything returned to normal for once, today we are back to having problems again. We cannot go on like this
1 Like
Do you want to do something or not?? I have to complete a project and by doing this I’m just wasting my time, this is all shameful and embarrassing, once I’m finished I will never use anything with Google in it again if things are like this
Hey All,
We apologize for the delayed response. Could you please confirm if you are still experiencing this error?
A 503 Overloaded error typically signifies that the service is temporarily overloaded or encountering a capacity constraint, often during periods of peak usage.
This status indicates a temporary capacity overload specifically within the Gemini 2.5 Flash infrastructure. We strongly advise implementing exponential backoff (retrying requests with progressively increasing delays) to effectively manage these transient failures. Should the instability persist, please consider temporarily redirecting your API calls to an alternative model.
Thank you!
But I understand what the problem is, and I don’t need solutions that only partially resolve the situation. The fact is that I want to put pressure on AI developers or anyone who can help solve the problem to increase the number of servers available so that at least the server overload is drastically reduced. that’s the problem, and I’m trying to make you understand that in every way possible, but if those who can do something don’t want to hear it or do it, then that’s another matter. If it’s a question of costs to support the whole thing to the detriment of consumers, then I understand, but it’s wrong. Those who pay, but also those who are entitled to use it freely, must be able to do so, otherwise it’s a joke. For a month, I never had any overload problems and everything worked perfectly. Now you want to explain to me why, all of a sudden, nothing works anymore at the same times it worked before? Something must have changed, and that’s why I’m insisting on making it clear that this is wrong, that it doesn’t work, and that many people will abandon it.
In my case, I can’t switch to another model because the Gemini 2.5 Flash Lite gives inconsistent responses to the topic, and it’s not what I need and what I paid for through a third-party service. I have a free API key for 250 calls per day, but if I have to request 10/20 or 30 calls to open an article, you can see that this is not sustainable for anyone, not just me.
1 Like