Question about Cursor AI premium model usage counter behavior

Discussion
1

I’ve noticed something interesting about how Cursor AI counts model usage, and I’m wondering if anyone else has observed this or can explain what’s happening.

I’m using the paid version of Cursor and primarily work with the Sonnet model. Looking at my usage counters:

Premium models:
19/500 (I’ve used 19 requests out of my 500 monthly fast requests quota)
gpt-4o-mini or cursor-small23: 23/No Limit

Here’s what’s puzzling me: Initially, I thought Cursor might be smart enough to route simpler queries to gpt-4o-mini or cursor-small23 automatically to optimize usage. I specifically remember using gpt-4o-mini once after signing up for the paid version.

However, I just made a new, relatively simple request with Sonnet selected as my model choice, and I noticed both counters increased simultaneously. This happened despite having Sonnet explicitly selected as my model.

Does anyone understand how this counting system works? Is this intended behavior, or am I missing something about how Cursor handles model selection and request counting?

2

Hey, when you select a model for your requests, you get a response from that model and generated code to make changes. This process is managed by a special model, which is considered an unlimited model in your quota.

1 Like
3

Thanks for explaining this! Just to help others understand: When using Cursor AI with Sonnet, each request actually involves two models working together - Sonnet handles the main response (counting towards the Premium quota), while a background model helps process code-related tasks (counting towards the unlimited quota). This explains why both counters increment simultaneously, even when only Sonnet is selected. Really helpful to understand how the system works.

1 Like