[503] This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later

Gemini API
1

Gemini 2.5 Pro has been returning this issue TIME & TIME again, its a huge issue when your using this as a system to run a business & clients expect the AI to work. “[503] This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.” Also I find it extremely upsetting and inconsiderate that you are not refunded if you have a workspace account for the usage due to this error, any tokens used are still charged to you so if you try to re run it 20 different times during different times in a day and it still isnt fixed you get charged 20 times when its their fault. I expect something to be done about this but knowing them it will be weeks before a fix is in place

14 Likes
2

Same here, often 5 retries before getting an answer…

1 Like
3

same issue I’m facing here.

5

same - I’ve been experiencing this for the past 24hours.

6

I can’t believe that Google can’t afford to add more capacity to solve this problem!

7

Been also experiencing this for the past few days…the service is effectively unusable.

I get that these things happen, but any kind of official communication would be tremendously helpful. The Gemini status page shows all green…which can’t be right given the number of reports? Even just a simple “we know there is a problem, we’re working on it” would be helpful.

I’m trying to figure out if I should just tough it out and wait for a resolution, or if I should switch to a different provider. The latter option would be non-trivial for me given how I’ve integrated Gemini into my project, but I don’t think I can afford these kinds of unexpected outages with no resolution in sight.

3 Likes
8

There are so many threads on this topic. The forum moderation is actively silencing the issue like in this discussion Handling 429 / 503 errors from the Gemini API - #20 by Albert_Sunjaya

1 Like
9

It makes me so mad cause I am paying for a product they can’t deliver

1 Like
10

Its been a full week and i’ve gotten almost all my prompts rejected, 2.5 pro is so inconsistent at this point its basically unusable, but 3.1 pro has a new problem that 3.0 didnt where its generated text is super robotic unless you stipulate, prompt by prompt, every time, for it not to be, its absurd. I just wish I knew when 2.5 pro would actually be available to use again.

1 Like
11

Same here, no idea when this could be resolved :disappointed_face:

I end up by setting up retry logic in my app, & it worked on 6th try which is very bad, i am using 2.5 flash

1 Like
12

same goes here, which is quite disappointing. The issue persists across gemini-3.1-flash-image-preview, and gemini-2.5-flash-image.

gemini-3-pro-image-preview does return an image sometimes, but it is significantly slower and much more expensive compared to the other models.

1 Like
13

I’m facing the same issue. Gemini keeps breaking and messing up my workflow. Tried 2.5 Pro, 2.5 Flash, latest Flash, even 3.1 Pro Preview… still hitting limits and errors.

1 Like
14

Limitation of AI or unprofessionalism? Why bother creating workflows and paying for APIs if suddenly the underlying model crashes ”due to high demand”? I had to update my gh workflow and add a higher number of attempts when errors appear, in the hope that the so-called automated workflow will function as expected. It’s either that we are long away from AGI or the "democratised access to intelligence” is a myth!

15

Nobody is responding to the issue. I posted at least 10 times about this issue. no response. just don’t understand who is getting a response about their issue. only users are commenting they have the same issue.

1 Like
16

Well…I do see moderators responding to issues not related to the 503 errors, which give me the impression that they’re aware of the issue but are ignoring it. And I’m guessing the reason they’re ignoring it is because its an issue they cannot resolve.

Rather than this being a technical issue, I’m guessing Google simply does not have the hardware to handle the demand, which isn’t something they’d like to advertise given that their hardware and computing power is probably they’re biggest selling point vs their competitors.

This is all speculation of course, but its really strange not seeing any kind of acknowledgement to the issue that several users are clearly dealing with.

1 Like
17

Does anyone find the solution for this issue?