Same Issue here. 503 only so the actual response never works
I don’t think it fixes the issue, as the problem comes from file upload, not gemini model availability
Thanks @daniel_lublinsky - that’s a clever workaround you’ve discovered! I’d prefer to not have to rearchitect our app just for this outage, as we currently upload the document once and then run several queries against it.
Maybe your suggestion will help others here, though, if they are better positioned to make quick changes to their codebase.
Hopefully someone at Google is seeing enough 503 errors to raise an alarm and investigate it 
Also wanted to say thanks to @user3473 for the comprehensive write-up! You’ve done the rest of us a big favour.
Same issue here. 503 error in file upload.
Thanks, for this was simple and moving from types.file.upload to types.Part.from_bytes
And pushing the bytes to the cache directly
I have raised the issue in their GitHub repo here: Issue · GitHub
Thanks @user3473 !
Same here! My users are fuming 
It’s working again, at least for me. Hope for you guys as well!
Seems like it’s working again!
seems that it’s working again
I guess its fixed now but from morning we had our production down, so I came up with the hotfix solution.
I’ve implemented a conditional routing solution to stabilize uploads:
If the uploaded PDF is ≤ 25 MB, the system processes it via inline upload requests.
The route inspects the payload size and errors like upload terminations from logs and dynamically decides whether to use inline or alternative upload paths.
For added flexibility, I recommend introducing a kill‑switch flag (boolean) in config. This allows ops/dev teams to toggle inline upload on/off at runtime without code changes useful for incident response or load management.
Same here! Hopefully they fix this soon, I’m stuck
I am still having this problem. Anyone else still having the issue?