IMO, o1 is vastly superior to GPT-4o. As I understand it, on the free cursor version, I can use my own openAI API key to use o1 limited only by the capacity available to me from openai. But on the paid version, I need to pay $0.4 for every request.
If my understanding is correct, this is actually penalizing the paying members.
I feel o1 should be free for paying cursor members. You could charge a different tier for this, if required, but the important thing is to have as many queries allowed as an openai paying subscriber has.
1 Like
OpenAI has discontinued the Model Terminology “o1-mini” in their API.
It is now called GPT-4o-mini so that is likely your problem. Change your custom model setup to match it. Review in your OpenAI account the models available to you and note the naming changes they just made a couple days ago.
Just wondering if you have a reference for this, I’m struggling to verify. I thought gpt-4o-mini was based on the 4o line, and o1-mini is separate? Has this definitely changed?
The docs at https://platform.openai.com/docs/models/ don’t seem to reflect that.
I meant o1-preview, not o1-mini. o1-preview is great, but we get charged 40 cents per query. Its better to just use our own openAI API key for that. But cursor’s paying customers dont have that option. They have to pay the 40 cents per query. That sucks.
2 Likes
Hey, you access different models from one place, and o1-preview is too pricey. Your $20 Pro subscription won’t even cover it. You’re already getting a lot of value, if you used your API key, you’d spend way more. Are you suggesting the Cursor team do charity work? Here’s a screenshot from Twitter:
First, you’ll need to be at least on Tier 3. Second, if you are, you can manually add the model you need by clicking “Add model”.
Got it. thanks.
1 Like
Sorry if I caused any confusion. I am a tier 3 user at openAI, so those were the models I added to my cursor profile. But I now understand that Cursor embeddings are not active if I use my open API keys, and define models for use in Cursor ? So since this post I saw my key dwas isabled anyway. Thats fine, 01-mini is nothing without the Cursor embbedings.
The truth is Claude 3.5 blows openAI out of the water right now. On Saturday Nov 9th, I blew thru 250 fast queries and wrote 2500 lines of the most complex code I have written in many years. My 10 or 12 hour session was so unbelievable … I am loving this environment like nothing else ever !!
I paid another 20 bucks for my Pro account upgrade to 1000 fast queries. I’m very happy with Cursor, I even activated the Shadow workspace, though I don’t know what the status is, if its working or not ?
Screenshot attached of what my GPT4 model selections are as of today …
1 Like
Hi knaig,
To clarify, o1-preview is the full GPT4 LLM, o1-mini is their scientific only small model, it’s meant to be used for coding tasks as opposed to o1-preview, which is slower and way more expensive than o1-mini if you look at the screenshot I attached, down thread, you can see the openAI selections I have with my account there.
Using my openAI API key I can access my openAI account from within Cursor and define those models in Cursor, and be billed to my openAI account, even if I use them here.
use the mini model not preview
https://openai.com/index/openai-o1-mini-advancing-cost-efficient-reasoning/