Cursor AI Pricing Explained
July 2025 update: the Cursor pricing model changes a lot and as I was expecting, what I wrote back in this article back in May 2025 is not 100% accurate anymore. However it seems like Cursor messed up big time with one of their recent June 2025 pricing changes -> for more details, see this official statement from Cursor: Clarifying Our Pricing and also What Happened To Cursor?
OK, back to what I wrote back in May 2025:
0. Introduction
The Cursor AI Code Editor is an impressive piece of technology to say the least… but what’s not so impressive is the way the Cursor pricing model is explained.
So if you’re confused with what completions, requests, slow requests, free requests, premium models, free models mean and how much they cost, you’re in the right place: I was initially confused too, which is why I’ve created this article to share my findings… you’re welcome!
Note: the Cursor pricing model seems to change a lot so if you’re reading this in the future, the monthly amounts I’m describing here may not be valid anymore… however the Cursor features I’m about to describe are unlikely to change.
1. Cursor Business Plans
Let’s start with the 3 plans: Free, Pro & Business.
Those come with a certain number of completions, requests and slow requests… what are those?
1.1. Completions
This is what gets consumed when you use Cursor tab autocomplete feature.
For the Pro and Business plans, you get unlimited completions, however for the Free plan, Cursor contradicts itself:
- The pricing page states that you get 200 completions/month (see screenshot above)
- The tab completion documentation states that you get 2000 suggestions/month (see screenshot below)
1.2. Requests
This is what gets consumed when you use Cursor chat feature in combination with a Premium model. That last part is important because Cursor also gives access to Free models that don’t consume requests (more on this later on).
There are 3 types of requests:
Type #1: Fast Requests for Premium models (see documentation)
- The Free plan includes 50/month but only when using the GPT 4.1 Premium model.
- The Pro & Business plans include 500/month (the Pro trial plan only includes 150 though).
- When you’ve consumed all of them, you start using Slow requests: you can enable usage-based pricing to continue using Fast requests.
Type #2: Fast Requests for Free Models
- The Free plan includes 500/month.
- The Pro & Business plans include an unlimited number.
- See one of the following chapter to understand what a “Free Model” is.
Type #3: Slow Requests (see documentation)
- The Free plan doesn’t mention how many are included.
- The Pro & Business plans include an unlimited number.
- You start using Slow requests when you’ve consumed all the Fast requests that are included by default in your plan.
- Slow requests are free… but they’re slower… hence their name… get it? :)
- The Cursor documentation also calls those free requests: this is confusing because you also have “Fast Requests for Free Models”
1.3. Usage-based pricing
When you’ve used all the Fast requests that are included in your plan, you can switch to usage-based pricing. This means that you can now use as many Fast requests as you want… but you pay for each one of those!
Get Laurent Kubaski’s stories in your inbox
Join Medium for free to get updates from this writer.
As I’m writing this, each additional Fast request will cost you $0.04: that’s exactly the price of the Fast requests that are included by default in the Pro plan ($20 for 500 requests = $0.04 per request).
1.4. Use your own API key
An alternative to paying a fixed price of $0.04 for every Fast request, is to use your own API keys: this way you pay the same price as if you were directly using the APIs of the various LLM providers like OpenAI or Anthropic.
If like me you like to use OpenRouter (this allows you to use the Open AI SDK with any model of your choice), here is how to configure the Cursor IDE:
The main problem with this approach is that this will prevent you from using a lot of the Cursor IDE features. For example, the agent mode doesn’t work if you use your own keys:
Anyways: after entering my own OpenRouter API key, I’ve enabled the “Ask mode” and then asked Cursor to explain what a basic function in my project was doing. As expected, I did not consume any Fast request, and I ended up paying $0.003 to OpenRouter for that.
Yes, $0.003 is way cheaper than the fixed Cursor Fast request price of $0.04 BUT in this case I only consumed a very limited number of tokens: for larger requests, a fixed price of $0.04 no matter how many tokens are used is certainly cheaper.
1.5. Premium models vs Free models
The number of Fast requests that get consumed depend on which model you use. For example on the screenshot below, you can see that Claude 3 Opus consumes 2.5 Fast requests for each interaction:
But wait: as you can see in the screenshot above, some models — like “Cursor Small” — do not consume any Fast request: these are Free models.
As I’m writing this, Cursor supports the following Free models: Cursor Small, Deepseek v3, Gemini 2.5 Flash, GPT-4o mini (limited to 500 requests/day with the Free plan), Grok 3 Mini Beta,
2. How To Use A Free Model
On the Cursor model page, find a free model, then hover your mouse on the copy-paste icon and click on it: this copies the internal name of the model (on the screenshot below it’s “cursor-small”) to the clipboard.
Then in the Cursor IDE, open the Cursor Settings page and if you don’t see your Free model in the list, click “Add model”, then paste the value you copied and hit “enter”: the Free model should now be in the list (make sure that the checkbox is selected in front of it).
Finally, in the Cursor IDE Chat window, open the model switcher, disable the “Auto” option and select your Free model in the list:
Additional notes:
- Obviously some of the Free models will not perform as well as the Premium models.
- AFAIK, Cursor doesn’t document if there is a limit on how many times you can use a Free model.
3. Tracking Usage
3.1. Settings page
The first location where you can track usage is the settings page:
- This page is not available when using the Free plan (it redirects to the dashboard page described in the next chapter)
- This page displays:
- how many Fast requests have been consumed by Premium models and by Free models. - This page does not display:
- how many Fast requests have been consumed by usage-based pricing.
- how many tab completions have been consumed.
3.2. Dashboard page
The second location where you can track usage is the dashboard:
- This page displays:
- how many Fast requests have been consumed by Premium models.
- when the number of Fast requests is reset.
- how many Fast requests have been consumed by usage-based pricing. - This page does not display:
- how many Fast requests have been consumed by Free models.
- how many completions have been consumed.
4. To Recap
Note: all numbers are “per month” and are as of May 31th, 2025.
+------------------------------+---------------------------------------+
| Feature | Free plan | Pro/Business plan |
+------------------------------+---------------------------------------+
| Tab completion | 200 or 2000 ? | unlimited |
| Fast request (Premium model) | 50 (with GPT 4.1) | 500 |
| Fast request (Free model) | 500 | unlimited. |
| Slow request | <not documented> | unlimited |
+------------------------------+---------------------------------------+
