Error: The model is overloaded

Gemini API
25

Same here. Does anyone know where to find the latest updates on this incident? I’ve switched all AI models to another platform for now and am waiting for the right time to switch back!

1 Like
27

Also dealing with hang-ups and 503 model is overloaded error. For Flash Latest & Flash

1 Like
28

In the same boat unfortunately. I was getting consistent, safe, results using the web UI/aistudio, so I started to make fetch calls to the https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash:generateContent endpoint and started getting the same 503 over and over again. No matter free or paid; random 503’s with the same payload. I am using a pretty basic setup with a responseSchema.

I was excited to try this out in production, but there is no way im pushing code to prod with a 40% 503 error rate lol.

1 Like
30

i am getting the same error using Genkit

[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-latest:generateContent: [503 Service Unavailable] The model is overloaded. Please try again later.
Error: [GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-latest:generateContent: [503 Service Unavailable] The model is overloaded. Please try again later.
1 Like
32

Also happening for me. What can we do to fix this??? My entire app relies on this API to scan and process images. I can afford for it to be offline like this. Using flutter, tried the official generative ai package from google and the flutter_gemini package. Nothing fixes it.

1 Like
33

Same thing here. Will Google not solve this problem !!!

1 Like
34

Same here:

“code”: 503,
“message”: “The model is overloaded. Please try again later.”,
“status”: “UNAVAILABLE”

1 Like
35

Same 503 :face_exhaling:
https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash:generateContent: [503 Service Unavailable] The model is overloaded. Please try again later.

1 Like
37

this is still going on its been happening off and one for a while now but recently its happening more and more and in the last hour every single request i have done is throwing that error does anyone have any updates or is there any thing i can do besides wait for a fix

39

I can confirm this is still happening. Been using gemini with no issues while developing a react native app. Today I started getting random 503 errors. Im forced to stop using this entirely until something changes.

2 Likes
40

Yeah, It just started doing that. For some reason maybe some people are rigorously now trying to use knowledge distillation to copy gemini now.

1 Like
41

did u find any fix? I m also facing the same problem from past few hours

42

Same problem from few ours . At first i thought it’s due to large request from my side . but Even after changing the new api key . 80% requests are failing

43

So one, I was not expecting this to get any replies or comments, but I’m glad it’s somewhat helpful. Here is what I did to help with the errors.

First off, this is a server issue from Google so we have no control over when we will get a 503 error. What we do have control over is what happens once we get the error.

In my react native app, I am using Gemini to come up with some titles for articles. Initially, I would just call the client In my anpp and wrap it in a trycatch. The issue with this is it would throw my app and everything after that Gemini client would NOT run because the Gemini client returned 503.

So since I can’t control if it’ll give me an error, what I did was before running the—Let’s just call it the “Important function”, I set up a test function to send a small prompt FIRST before even starting the “Important function”. This is to verify if Gemini is even working. Let’s call it “Ping”.

Now, once “Ping” actually returns back a response, then I can run the “Important function”.

The moral of the story is, We are using services that we have no control over if we’re being honest, but we can control the edge cases. It’s a little bit extra work to have to verify if Geminis servers are working, but it will allow you to better customize situations where Gemini is off/not available.

For example, you could test if Gemini is working and if it is then continue with the rest of the stuff in your functions/app
if not,
Maybe you can choose another AI or implement in the moment to take its place. I’m sure there’s some official technique for whatever this is called. This is just more my way of dealing with things that can be unreliable.

I hope this isn’t too long, but I tried to explain this the best way that I can. Let me know if you have any questions :smiley:

44

I think the best current option is to migrate to using OpenAI’s library, since Gemini is now compatible with it, and use other models until Google fixes its servers.

1 Like
45

Thanks a lot for taking the time to reply to my query. I have one last question. You mentioned that it’s an error from Google’s servers, right? How long did it take for your 503 errors to get fixed – like 1 day or a week?

1 Like
47

i was also facing the same issue for past 6 hours but finally it resolve now, I get the data from the model (gemini-2.0-flash-exp)…

2 Likes
49

It is sporadic, not a total failure. Just low reliability seems to come and go.

1 Like
50

I am also getting the below error from gemini-1.5-flash

[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash:generateContent: [503 Service Unavailable] The model is overloaded. Please try again later.

I switched to flash-8b for now and it is working, but this is superfrustrating, because the reliability is very less

51

I am also facing the same problem for 1 hour. I have made a chatbot by Gemini’s api. I dont understand how to fix it.