I tried gemini-3-pro and flash models, but only flash model gave me 503 errors frequently. This has been happening consistently and is still occurring today.
Same here getting 503 for almost two weeks now.
Is there an ETA for a fix?
I have the same issue. Is there any solution?
Previously, I thought that Gemini 3.1pro was in the process of being launched and was undergoing testing, so there might be delays or glitches. But it seems that this is not the case. At a certain specific time period, Gemini 3, 3.1pro, Nano Banana would start to randomly return 503 or “Failed to fetch”, but AI Studio was not affected. I don’t understand what exactly “a large number of requests” suddenly appeared and occupied the server’s resources.
2 Likes
From what I saw, Google discontinued several routes that now point to the gemini-3-preview. That is probably the cause of so many problems. I’m still having this issue generating images with Nano Banana, but I saw that on platforms like Kie.ia, Nano Banana seems to be working, which makes me wonder why it’s not working on Google’s own API.
1 Like
I have the same problem; no matter when I use Nano Banana, I get error 503.
Gemini CLI user here, with company API Key in Tier 3.
Gemini CLI has been mostly unusable for the past 2-3 weeks. You get constant “503” or “fetch failed” errors. This is happening mostly with 3.1 Pro model, but “Auto (Gemini 3)” works just as bad. Selecting 3-Flash manually seems to be working better.
<<The team is very aware of this and fighting hard to shuffle around capacity>> one month ago does not seem like the most appropriate response we should get, given the scale of the issue. Additionally, the Status page says everything works as a charm, so you cannot even trust that.
8 Likes
I’m both surprised and not surprised that I am experiencing this same instability, then finding and reading this thread that was posted over a month ago about this issue. Google, this is not acceptable for a model that is presumably intended to be used in production environments!
1 Like
Still getting these errors on paid api keys… on gemini 3 pro preview and 3.1 pro. Google has been shuffling around capacity for over a month hahaha…
I was bitin gmy nails thinking it was only me. The Pro 3.1 is giving me internal errors. I couldnt rely on AI studio for the past couple of days and its really shocking to see how negligent Google has been recently. Stable 2.5 is way better than unstable 3.1.
keep encountering 503 errors, any solutions?
```
{‘error’: {‘code’: 503, ‘message’: ‘This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.’, ‘status’: ‘UNAVAILABLE’}}
```
1 Like
503 errors all the times with gemini 3.1 pro
This should be unacceptable coming from Google and an 11-day-old model. Literally, there is no way you could do anything with the gemini-3.1-pro-preview model at this point for what it feels almost a week now. We get that it is a “preview” model, but this is commercialized as a model that could, and should be used in production since it’s out in the open and available in your API suite from day one. Do better
Same here - `Auto (Gemini 3)` is unusable.
Can we at least have a status page that provides useful info, e.g. not all green when it is pretty obvious there is an outage.
2 Likes
textPayload: “[Gemini] gemini-3.1-pro-preview → 503. Wait 3s (5 left)”
503 UNAVAILABLE. {‘error’: {‘code’: 503, ‘message’: ‘This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.’, ‘status’: ‘UNAVAILABLE’}}
1 Like
Hello,
the service isn’t usable. Many of my use case require many request and regulary gemini doesn’t work for a full hour and my process stop. How can I get refund ?
It’s truly impressive how Gemini API Status page maintains a perfect “All systems operational” while everyone here has been dealing with constant 503 errors for hours. Perhaps the real AI breakthrough is the hallucination of a fully functional service. 
7 Likes
Me too. Just now
gemini-2.5-flash-lite,
ApiError: {"error":{"code":503,"message":"This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later.","status":"UNAVAILABLE"}}