Gemini 3.1 Pro has become virtually unusable overnight. Every query gets me the same response (“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”), whilst other models work just fine. It’s a pity, since the model is actually really good.
Anyone else having the same issue?
Same problem here, very annoying
The first day Gemini 3.1 pro available I built up some code with it and works well
but keep connection error today forced me to switch back to gpt codex
It’s on Googles side, they are fumbling 3.1 for like a week now. Cannot even access 3.0 in their CLI..
Unfortunately the retries cost money (I pay per request).
Hey, you’re not the only one. Over the past few days, several users have reported the same thing with Gemini 3.1 Pro. It looks like an upstream issue on Google’s side, since other models work fine for everyone.
Related thread: Bug Agent stuck on “Taking longer than expected” with Gemini 3.1 Pro
While Gemini is unstable, you can switch to another model. Auto works well as a fallback since it picks an available endpoint automatically. The team is aware of the issue.
A couple quick things to check, just in case:
Ctrl+, > search for “HTTP/2” > enable “Disable HTTP/2” > restart Cursor.Let me know if other models start failing too.