Fast Request Pricing Transparency

Cursor Team,

Given the discussions around issues with slow requests, can you create 100% transparency around pricing for fast requests.

What would really hep is a quick calculator (or at least a look up table) in this style: OpenAI API Pricing Calculator | gptforwork.com

As I understand, there are 3 ways to pay for (500) additional large context Claude Sonnet 3.5 requests:

  1. Settings | Cursor - The AI Code Editor → add fast requests at $20/500 requests
  2. Settings | Cursor - The AI Code Editor → enable usage-based pricing at $100/500 requests ($0.2 per request)
  3. Anthropic API Key (in IDE settings) → your colleagues estimate in previous post $15.00/500 requests

Can you please clarify if my pricing is correct?

Thanks

Hey, you are almost there, but number 2 is much cheaper!

For Claude 3.5 Sonnet, the requests are priced the same as normal ($0.04/request), billed each month. If you use 500 requests, that works out to the same $20 price as normal.

While the API Key may be cheaper directly, usage-billing with us means you don’t have to manage API keys, have billing set up with Anthropic, and help support the development of Cursor too!

Hello,

So be be clear, when using user-based pricing, you charge a margin on top of what I would pay using the Anthropic API Key directly?

What is the margin you charge on top if that is correct?

Thanks,
Andres

I believe the rough average for a request direct to Anthropic would be ~$0.03, although this could fluctuate.
For usage-requests direct to us, we charge a flat rate of $0.04 per request.

What does it mean that usage-based pricing is billed for every 500 requests (as mentioned on your help page)? Does it mean there is a minimum of $20 each month for usage, or will billing occur once it hits 500 requests?

Does usage-based pricing also apply to GPT-4o?

Thank You

Hey!

Claude 3.5 Sonnet, GPT-4o and GPT-4 are all priced at $0.04/request.

By default, you are charged for your usage pricing at the end of the month.
However, if you hit $20 in usage, we will bill that immediately!
But you can definitely spend less than $20 in usage pricing.

Hey Dan. Could we please get this info in a official format in the Usage section of account settings: Settings | Cursor - The AI Code Editor?

With all of the new models being added, I have no idea what’s counted as premium or not and it appears the log of each use is now gone. This is not transparent at all. I saw in another thread that you mentioned deepseek-v3 is not premium, i.e. always fast with no limits – that’s something your customers would highly benefit from knowing.

Hey, you can find the pricing details on this page:

Killer, thanks. I’ve been using Cursor for quite awhile and had no idea this was here. Would still suggest updating the out of date information in Settings | Cursor - The AI Code Editor and linking to this documentation.

There’s also conflicting information. Is gpt-4o-mini free or premium with pro plan? Needs some work.

It’s free, sorry for the confusion!

We are hoping to improve our documentation very soon, so appreciate the notion on this!

1 Like