You can use this model with your API key or without it in Max Mode. If you don’t have this model, just add it manually.
I did add the model gemini-2.5-pro-preview-06-05 and it says doesn’t have good agent support yet. Were going back to the same issue we had at the beginning of this post.
the issue was fixed for gemini-2.5-pro-preview-05-06 however on the latest June model were running right back into where we started now.
MAX MODE also gets disabled when you enable it on gemini-2.5-pro-preview-06-05 here is a video demonstrating how max mode is working on the 05-06 model but NOT on the 06-05 model. This needs to be configured on Cursors end. @danperks fixed this last time but its not a solution if this has to be fixed every time a new gemini model comes out.
I meant the model name gemini-2.5-pro.
gemini-2.5-pro model CANNOT be used when using your own API key, it doesn’t let you add it. That model gemini-2.5-pro can only be used when API key is OFF and you are using it directly through cursor.
I gave you videos and screenshots, have you tried experimenting yourself first? Maybe show me a video of you getting is to successfully work with your API key enabled on the gemini-2.5-pro Or gemini-2.5-pro-preview-06-05 model
@danperks is fully aware what the original issue was and he had it fixed. I suggest you coordinate with him on this task.
Hi there, That’s not my experience (on v1.0.1): since the fix went in ~26 days ago, I’ve been using the existing (i.e., I didn’t have to add) “gemini-2.5-pro” model, using my google api key, with MAX mode enabled.
It’s correctly marking those requests as consuming 0 requests.
As mentioned, “gemini-2.5-pro” is “preview-06-05”.
What model do you see in the chat for the model selection, can you share a screenshot?
And can you also confirm you have the api keys toggle enabled
@shaggy2626 You can just use the standard Gemini model names in Cursor when using an API key, and we handle the rest.
Adding custom models is only used for cases where the model is not already available via Cursor.
Hi Dan,
The standard model name gemini-2.5-pro is not visible when I enable the api keys so I added it gemini-2.5-pro manually and I’m getting this error.
You’ve got a lot of custom models there. Perhaps there’s a paging issue?
I believe when you disable api keys, those custom models disappear from the list, so maybe that’s why you see ‘gemini-2.5-pro’ re-appear when you disable your Google API key.
If you view all models or search does it appear?
Also, try disabling all API keys except for Google, as there might be a conflict.
@deanrie you know what, that actually worked. Disabling all api keys except google showed the 2.5 pro model however its still using the gemini-2.5-pro-preview-05-06 rather than 06-05 while I understand gemini says they are “pointing” it 06-05 I definitely see a difference in the output as 05-06 seems to be very chatty vs 06-05 and I also I noticed When using your own api key with the gemini-2.5-pro you don’t get the “planning next steps” thinking output.
I’m constantly switching between models during a convo which is why I keep openai, claude and google api keys all toggled on which clearly causes a conflict. Possible its a bug?
Screenshot still shows that’s its using the 05-06 model on the cursor dashboard
well for temporarty patch I modified the /resources/app/out/vs/workbench/workbench.desktop.main.js file and now I’m able to sucessully use gemini-2.5-pro-preview-06-05 with max and thinking mode with no issue. I get the full thinking output. Using my own API key, exactly what I needed.
