I notice that more requests are being used than simply one per message in the composer. Does Cursor use multiple requests out of the 500 monthly premium requests? I couldn’t find information regarding this on the pricing page.
Hey, yes, if you’re using the claude-3.7-sonnet-thinking model, it costs as 2 requests. We announced this in the previous version.
Thank you for the answer. Would you also be able to answer the other two questions? I’ll reiterate, and also rephrase, them:
This is from my own experience, so take it with a grain of salt.
Submitting an Agent prompt with up to 25 tool calls consumes one request. If your request extends beyond 25 tool calls, Cursor will ask if you’d like to continue which will consume a second request. – Cursor – Models
They cost the same, so long as you aren’t using thinking. The link above shows the breakdown at the top.
You can continue to use 3.5/3.7/3.7t even after your fast credits are gone.
Thank you. I appreciate it. Hopefully, a community member such as deanrie – or a staff member – can come and verify these statements, which would be appreciated.
Nevertheless, thanks for the help. It’s quite reassuring.
One thing I’m wondering: after your 500 premium requests are over, you can still use Claude 3.7? Does it just mean that it takes longer? (And would it also trigger the high demand error that free users get?)
And if you’ve used your 500 requests, can you no longer use ‘thinking’ (reasoning) mode?
Hi, i would like to add to previous poster info:
I mean, it’s in the documentation I shared what they allow. Plus, read #3. I use Cursor 12+ hours a day, 7 days a week. I run through my credits on the first day after they roll over and use slow credits for the rest of the month.
Thank you. I have noticed, however, that it seems to cost 3 requests when I use Claude 3.7 Sonnet with thinking mode enabled. After a while of not using Cursor, I sent one message in the composer, and my requests for premium models went from 139 to 142, instead of 141.
Do you know what could be causing this? It did not exceed 25 tool calls. I wasn’t using the Max version.
Note that the list is not always updated immediately so make sure to check the timing.
For me it doesnt use any extra counts on any model.
Ah, yes, that’s likely it. Thank you.
Can one enable the thinking and large context modes at no extra cost once your 500 monthly premium fast requests are up? I apologize for the numerous questions, but this has been unclear to me. I have read the docs but couldn’t find an answer to this question.
No worries. its good to get good understanding.
Personally I have not tried Thinking and Long Context on slow requests.
The docu shows the pricing on Usage Based pricing and it would cost extra for Thinking and Long Context.
No worries. That’s what the forums are for. 
Yes, I mull through my credits in two days or less most months and use slow credits the rest of the month unless I have a project for a client and need it done ASAP; I’ll stack the $20 subscriptions and then reduce them before it renews. You really won’t find a better deal with 3.5/3.7/t.
I haven’t noticed any issues, and I leave 3.7/t on all the time, which is all I like using most of the time. I haven’t tested my prompt system with V3 or 2.5 Pro yet. That’s next on my list. The system I built for Roo works amazingly well with 2.5 Pro, except when it gets stuck in a loop and won’t research as instructed. Hoping I have similar results with Cursor.
