Hey, thanks for the report. Gemini 3.0 Pro launched on Nov 18 and there have been some initial provider issues that the team is currently working on.
Please try the following:
Create a new Agent chat (don’t switch models mid-conversation)
Test with a simple prompt to see if the overload errors still appear
If the issue persists, please share:
The exact error message or a screenshot showing the “provider overloaded” error
Request IDs from the failed attempts: Chat menu (top right) → Copy Request ID
1–2 example prompts that trigger the error
In my situation, it messed up my git commits and damaged my database it was working on via MCP - accidentally re-running any prompt that’s already been partly run is absolutely going to be disastrous in most situations.
I upgraded to the $200 Cursor level, but I still have been unable to get any work done for almost two days. All of the good models fail with an “overload” message. Even “Auto” started to fail. Does anyone have any idea whether this is short-term, or should I start looking at Antigravity or Codex instead?
Steps to Reproduce
Try to do any complex feature implementation.
Expected Behavior
Product works as advertised – tasks complete without back end failure.
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
Don’t bother Antigravity. It is even worse. And at least I have not been able to find place to put my own gemini API key there, so that path is blocked too. Cursor has at least place for the key even tho it is broken at the moment.
But overload started appearing with own key too, so I don’t think there is anything for users to do. Gemini3 is overloaded on google servers and it is going to stay like that until someone makes better model or they increase hardware.
The current issues are only happening because everyone is trying to use the model at the same time. It doesn’t matter what platform you’re on—access is limited right now. At least Antigravity is able to get some access to it and it if fails you can always prompt to continue.
