Hey I was wondering if anyone uses their own OpenAI API key and how much they pay if they use cursor regularly? I bought the $20 a month package and ran through my GPT 4 fast requests pretty quickly and am stuck with slow requests (which is still great) but I’m considering just paying for the API but I’m struggling to find out a rough cost it would come to for someone who regularly uses cursor.
Also when inputting a OpenAI key, it doesn’t even let you choose ChatGPT 4, only ChatGPT 3.5 options etc. Do you need to add a payment method first on OpenAI site to get access to ChatGPT 4 API?
How much/how many queries are you running a day and is that $7 using the ChatGPT 4 model? That is a very decent price indeed, I’m just worried about how fast it will add up when I’m constantly submitting my codebase when using chat etc, I’ve probably used over 1500 queries at this point whilst having it for nearly two days, a lot of which included my codebase so that’s why I’m worried.
Will give it a test in a couples day time anyway and see how fast credits rinse.
According to OpenAI’s usage dashboard, last month I made 90 requests and used about 250,000 tokens on GPT-4.
Your 1500 queries might not be measured the same way as my 90 requests. To get a better sense of the price, I would use the API for a day or so and then check your usage on the dashboard.
EDIT: Hovering the title of the usage dashboard says “data may be delayed up to 5 minutes”. I just verified that timeframe is roughly correct. So you could estimate your usage by running a @codebase query, waiting 5 minutes, checking the dashboard, and multiplying the result by how many queries you expect to make.
That’s a sweet idea, I was going to do something similar like monitor for 5 minutes with how many tokens I end up using but with my query amount maybe a hour is better like you suggested
With the amount I can see myself using, thank jesus for cursor.sh, even with slow requests it’s manageable and the only time I found myself in queue waiting was when I was using it at busy USA hours last night, today has been smooth just using slow calls.
I use my API with cursor heavily and I pay roughly 1 usd to 4 usd a day.
When you have codebase related queries, it will cost you more than simple queries. Also, it also depends on how you use. For example, if you select specific code and ask for modification will charge you less but when you ctrl a all code and ask to change a tiny thing will cost you more.
So far, I believe that is worth it. Consider paying that to somone… would they work for you that fast on-demand for that payout?
I agree it’s definitely worth it, I haven’t even got around to testing/using the API because the built in requests seem to be fast enough in general for me:)
While using the GPT-4 API, I swiftly realized that the Cursor Pro version offers exceptional value. My costs on the OpenAI API surged to $6 in a few hours of use, likely due to my extensive use of the Docs features, which resulted in a high consumption of tokens.
This is exactly what I was worried about, and why I also hadn’t got round to trying it with the API on pay as you go yet. Cursor is indeed exceptional value and I hope they can continue to provide unlimited “slow” GPT-4 requests in the future as it’s been amazing to use.
Get the OpenAI account and just use GPT-4o-mini and GPT-3.5 models. They are dirt cheap and that gives you something else to fall back on when the Anthropic model stalls out on occaision. Plus you can get past the rate limiting cursor has if you are a heavy user.
