Now the new gemini 2.0 PRO exp model has been released.
Testing it right now but it seems to be the first model to beat Sonnet 3.6 in frontend.
i mean we have 3.6 since 3 months and no one was even close to it so maybe finally we have a competitior.
4o still ■■■■■
Unfortunately, the pro version on Google AI Studio gives me much better results than the API model connected to Cursor.
The new Sonnet Killer™
we will see haha right now no one has as good Agentic performence as Sonnet. The AIstudio variant of Gemini Pro 2.0 is very very good. The API version is okay but still behind Sonnet. Hope it will change soon 
Is this a premium model? (It seems yes on my end)
right now it is not available. You need to insert your google API key and than add the Model in the Model section. So its not a premium model and using your fast request atm
These gemini models aren’t a great fit with Cursor right now, at least with “apply”.
You get weird stuff in the output like (analysis.qmd is the active file):
— a/analysis/analysis.qmd
+++ b/analysis/analysis.qmd
@@ -1143,18 +1143,4 @@
Hmm i had this issue some weeks ago when the models where new. But tried it again and it works pretty good tbh for me. in other AI IDE´s for example all of those models are horrible. o3, R1, V3 are really useless they cant use edit tools
Potential is good, but getting too much issues with editing/applying. Hope it will be fixed
Same there is a lot of potential - especially with flash 2 because it is much faster and cheaper for many small tasks, but many times apply doesn’t work it seems like
I actually just reported this as a bug - though I kind of think its a bit harsh as its an experimental model. Gemini 2 Pro is low-key kind of awesome though. The context window is hard to beat for pumping in loads of Doc’s into your queries.