Gemini-2.5-05-06 is a G.O.D

Discussions
21

Yes check your current level and the spending minimum to advance to next level.

22

I just used it to help me on some issues and was pretty impressed. It was interesting about the code snippet in chat. I have a specific rule that says: “The user is a hobbyist and not a pro. never show code snippets to the user as a means of exposition or explanation. They are meaningless to him.” And lo and behold it doesn’t show me code snippets when it’s trying to talk to me and it doesn’t mess up as much.

1 Like
23

Ha! As a quasi-intelligent being I find your message leaving more impact on me, I must try it

24

I don’t know on what type of project you could refactor it in it’s entirety without breaking everything. I tried to refactor just a part of a 70.000 line project and it failed miserably, also in MAX mode. Claude 3.7 sonnet still seems to be the better option for me, even though Gemini-2.5-05-06 is statistically better at coding.

1 Like
25

All i saw was teh title of this thread. And it was enough to convince me. I’m glad to say im not dsissapointed

1 Like
26

I wish the pricing was listed and the request quota was listed in the app. Also, why do I have to press control± or whatever to change the model instead of auto?

27

Yes! Same situation for me — when I use gemini-2.5-pro-preview-05-06 (MAX mode), it completely stops using tools to update files. Switching back to gemini-2.5-pro-exp-03-25 (also in MAX mode) works fine again(Mac, using version 0.50.4)

28

Gemini Flash is not adopted for agent using at all. Gemini Pro is better but stuck often loosing file access and failing with edits much more often than Sonnet or OpenAI models. Hardly usable in Cursor but very good models itself

29

I rebuild it from the ground up from fresh PRD based on original version of the software. Now it uses SQlite as database, the rest is pure python.
Also in the same range 70k-100k lines of code.

30

The trend that seems to appear is that I’m using the gemini model via API of google (and thus paying extra) and most people here are using it via cursor causing it not to behave as good?

31

I’m new to trying to configure the exact model used by Cursor. When the headline says “Gemini-2.5-05-06”, is this different from the “gemini-2.5-pro-preview-05-06” checkbox option I see in Cursor->Settings…->Models ?

If it is different, how do I enable the first one (non-preview or “Gemini-2.5-05-06” this topic headlines)?

Is it only available if I setup Google/Gemini account billing? I only have a Cursor account at the moment. Otherwise, how do I add it to my available settings…->Models?

edit: I’m on Cursor 0.50.4, Darwin MacOS, Sonoma 14.0

33

sorry I abbreviated headline

34

so you are liking “gemini-2.5-pro-preview-05-06” is what you are saying?

35

Yea it works amazing with this autonomous workflow, just used gemini to make an update to the existing system, seems to work flawlessly so far. Removing my need to keep a changelog.txt

36

Is there something wrong with having code snippets printed in the chat? Does this add to the context and thus create worse responses in code files?

37

Maybe its fixed but this used to cause my gemini to start editing the file with the snippet and removing all rest of the code…

1 Like
38

Good to know, I’ll keep an eye out. First time trying Gemini today.

39

I believe they just dropped actual 2.5 pro (not preview anymore)

40

no 3.7 max is also good…gemini made useless changes for me

41

Noooooo, nonononononoon, new model is stupid. hallucinating, and it loosing track o conversation, and dont remember what we were talkign about, forgetting about files ive sent him, confusing posts with articles. this is much worse.

CURSOR TEAM, PLEASE I WANT THE OLD MODEL BACK!