[PARTIALLY SOLVED] Gemini models overloading with token windows of less than 20?

I’ve been using Gemini 1.5 Flash for quite a while now with very few lag, and maybe a few isolated errors. Starting yesterday, I’ve been getting many 503 Errors, Model is Overloaded immediately after sending prompts. The API now only responds less than 25% of the time.

Is there any solution to fix this error?

I have the same issue for about 15 hours. Also I think around less than 40% of requests is successful. It’s returning “503 - service unavailable” sometimes, so I think this is 100% issue on Google side.

Not exactly sure, but probably this is because of new Gemini version, which was dropped recently :rofl:

facing same issue on all gemini modal i tried my self learning project is stopped due to this.

Same issue, I think something is wrong on Google side

You are right man, my whole user base is irritated sie to this. I am also facing same problem for last 15-20 hrs.
Does anyone contacted gemini ai support??

Same here. If anybody finds a fix please post it here

Gemini changelog says the most recent update was the Gemini iPhone app, I don’t think this is directly related to the issue, though it is possible that it is indirectly overloading Gemini’s system. :face_with_diagonal_mouth:

[IMPORTANT]

I think I found the fix!

Try using Gemini’s experimental model “gemini-exp-1114”. I believe that Google allocated too many resources to the experimental model, possibly due to an imminent update.

Please tell me if you have any issues with the experimental model.

Does this model also have the same free tier as flash-1.5? Like 1500 RPD

No, unfortunately, it has a 50 RPD Limit like other Pro models.

for some reason i have max usage but i can still use my application…?

I have the same (quota reached), so the fix did not help :frowning:

I used this in my project andI have a interview and just now this is happening🥲

I think now all good guys

at least last 5 hours

I have the same problem. Since Friday evening I’ve been getting a 503 error message all the time.