When using Gemini 2.5 flash, I constantly get the message:
“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.
(Request ID: 882a5ae6-8b0a-44f4-bbd6-a9415ef6a0e0)”
This has been happening for weeks/months, not just recently. But I have been using gemini 2.5 flash more recently and it has just become unusable, as almost half of the time I submit a prompt it does that. I am on cursor enterprise if that matters.
Steps to Reproduce
Try to chat with gemini, a huge portion of the time it will return with that message.
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
I’m just gonna say this, because it’s the truth:
They do not care about any issues regarding Gemini, and at this point I suspect it has to be intentional.
There are plenty of threads about Gemini specifically having issues, for a long time it just straight up did not work at all - and to this day Gemini regularly has issues such as:
Not being able to properly apply edits
Not being able to verify having actually applied the edits
Excessive tool calling
Looping tool calls
Reduced performance
I can predict what is gonna happen here for you:
A: “I have forwarded this to the team and we will look into it” hint: They will not
B: “Unfortunately, this is on Googles end and not ours.” Sure
C: They will just ignore this.
Crazy how the Cursor team just can’t seem to get Gemini working properly after, I think it’s at least 6 months? More?
At this point I believe it is intentional for some kind of reason,
perhaps because you can get a Gemini API key for free and that would discourage from getting their higher cost plans.
If not, they are incompetent and simply can’t get Gemini working after half a year.
I’d love to just use Gemini 2.5 Thinking and Flash, but yeah.
Honestly, I just hammer AutoMode a lot because while not good it is more stable and I don’t have to think about pushing above my payment plan (so ofc they are taking that away also)
I’ve come to accept that cursor is just not as good as it was.
Ton of people will gaslight that its better than ever etc, but truth is that while we have gotten all the fancy new models and jidda jadda, Cursor was still providing far better result using Claude 3.7 or Gemini 2.0 in the pre 1.0 versions 0.47 and down.
I’ve learnt to simply code more and rely on Cursor less, it’s just a trap at this point when you might feel like you are saving time but you are not.
For every day it works, any day it doesn’t you lose not only time but momentum, energy and motivation. Not to mention all the frustration - like, even writing here and keeping up with whatever random dumb move the cursor team is up to is a waste of time.
Also, I’m not on the request based plan,
And I’m most likely out once they dump the unlimited Auto-use,
and it will probably feel great to not have to think about whatever this dev team is doing anymore and just have my energy go to my actual project.