What are fast and slow uses?

Discussion
1

Hello,

I am referring to this page:

https://www.cursor.com/pricing

And it doesn’t seem to specifically answer the question:

What are fast and slow uses?

Fast uses of premium models are given first priority by our backend. On Pro, once you hit your fast usage limit, you can still use premium models, but your requests may be queued behind others at times of high load.

At first, I thought Fast might be synonymous with Premium models, ie:

GPT-4, GPT-4o, and Claude 3.5 Sonnet

But then I saw terms like:

Slow Premium
Fast Premium

So I am interested to know what, exactly, fast and slow uses refers to.

And if we can we toggle fast and slow modes in the Cursor settings?

Below are some guesses for a more precise definition of what fast and slow uses are:

A) Fast and slow uses refers to the amount of compute power that is assigned to complete your request - fast uses are assigned more compute power than slow uses. Each prompt is equal to 1 use.

Or:

B) Fast and slow uses refers to the priority of your request in our request queue - fast uses are 1st priority, slow uses are 2nd priority. Each prompt is equal to 1 use.

Thank You.

(FYI - I seem to have gone through 500 fast uses after 2 days of doing fairly basic ‘playing around’ with Cursor to see what it can do etc)

2

An additional point I would appreciate clarification on:

Do the terms fast and slow only apply to premium models?

Are non-premium requests all processed at the same speed?

I hit my 500 fast use cap after a few days of pretty simple testing of Cursor and it’s functionality.

For me, these ‘slow’ premium requests are unbearably slow, after experiencing the fast premium requests.

I can pretty much chat all day, every day, about code with Chat GPT Plus membership.

So I am pretty used to that speed now, and everything else seems lacking.

Just providing honest user experience for feedback, I love Cursor otherwise.

3

From what I understood, fast and slow apply to premium models but for non-premium models they always have fast requests for pro plan, same speed as fast premium

1 Like
4

Non-premium models are always “fast” - they are just much cheaper and it probably the math adds up for the Cursor devs. The fast and slow apply only for the premium models - it’s probably kinda arbitrary but I guess they did some calculations based on average conversation length and it’s where they turn some profit. The unlimited uses are there just in case but I wouldn’t really rely on that if I needed to get some work done.
As far as I know you can subscribe twice and you get 1000 fast uses or if you want to you can provide your key and after the fast uses have run out you can switch to the LLM api for as much use as you need.
You’re probably doing a lot of small chats which wastes your included uses. Have some alternative subscription for brainstorming and fleshing out ideas and then use Cursor mainly for larger generations / chat with codebase which would be difficult to replicate in ChatGPT/Claude because you would need to copy a lot of files.

1 Like
5

For reference, just found an older post from Cursor which adds some clarification:

https://forum.cursor.com/t/what-exactly-are-fast-requests/46/2

2 Likes