I been going back and forth and if you take step by step you can make complex systems. it’s so good just a little tedious
yes, better than claude 3.7 max thinking.
fun fact, even with the error you can still add it, just put the toggle to ON and you can use gemini api key models. You can verify this by looking at the usage history. But its annoying and confusing and they shouldve fixed it a long time ago…
It will be fixed.
Its primarily because Gemini 2.5 pro doesnt know how to use Cursor, and vice versa. The Cursor team has Sonnet 3.7 implemented before it’s release, and have exclusive access to MAX.
I’m not a fan of Google, but if you use this model in Roo Code - it’s free, and does a much better job than Expensive Sonnet in Cursor. It doesnt run away with the faeries like Sonnet is known to 
. The 1m context window is amazing too. Yeah, I used previous Gemini models and was VERY disappointed every time. Always chose to pay over use them… now, choice isnt as simple.
Also, Cursor must consider how to implement it. If they put it native into Agent mode, people will complain about paying or just put their AI Studio API key in… and you can guarantee Google won’t incentivize it as they have competing products.
P.s. rumour has it 3.7 Opus or possibly even 4.0 Opus (just Opus) is coming very soon. That’ll shake everything up, as they won’t charge much more than current Sonnet, as it has some pretty worthy competitors now.
Are we being charged fast requests for a free model?
No its not a free model. Its a premium model.
gemini-2.5-pro-exp Google ✓ ✓ $0.04
paying premium for a recently free api ? ok, will try out alternatives to cursor.
This is the latest top of the charts model. What are you talkiung about free API?
For more than just limited usage (which means Cursor) a pay-as-you-go pricing model applies, with costs associated based on usage beyond the free limits.
Yep Cursor are not within the 100 requests a day free tier
I’m not aware of any public pricing, but I would expect it to be cheaper than Claude 3.7 and it would be great if Cursor reflected that (like they do with 03-mini) in the credit usage.
The pricing for Gemini calls makes no sense. Gemini is 1/30 the price of Claude in most cases.
Please, quickly incorporate this new model into Cursor
Did you check it? its already in Cursor and working
I used Gemini 2.5 in Agent mode and it won’t create files or folders for me. Sonnet 3.7 works fine. Is this due to Gemini not adapted for Agent mode by cursor or am I doing something wrong?
No it should work as Cursor team has implemented the Agent mode for the model yesterday and released it for all.
Add please your Cursor version and if you use an API key or not.
Version: 0.48.2
I tried using both API Key and directly the Cursor provided model in Agent mode but it wont actually create files or write code for me. It just shows everything in the chat window and shows it has completed the task.
Is it working for you?
Yes i use it all day.
Which mode are you in? Agent should work fine. In Ask mode it wouldnt edit the file.
I am using Agent mode only. are you using API key or the one provided by Cursor?
I never had the API key. just the regular one. I suggest adding the Request ID or submitting a feedback in the chat directly.
Thanks. I will contact Cursor. I kept thinking it is yet to be fixed by Cursor and gave up using it much.