How to add a custom model like DeepSeek-V3 (which is OpenAI compatible)

You know… You know you know… It’s the best there is and relatively speaking (Claude) it costs you a penny on the dollar. OpenAI just started a “free” competitor… Please don’t do that to us. I already know it will be throttled to promote their premium$$$.

I think I saw that. And it’s causing people to think deepseek is slow. Fireworks is slow. Deepseek is very fast, very very very accurate, and relatively speaking, 10^2 cheaper.

1 Like

How do people say: if it’s free you’re the product :wink:

Deepseek is Chinese and they always were popular for data collection (look at TikTok). So be careful with it, you never know what they do with this data behind the scenes. Obviously we don’t know what other West providers do with the data, but somehow I trust West more. Ideally all models have to be run locally for the best privacy.

Just sayin. I’m using it myself and I love it, but I tend to avoid sharing any sensitive data to it. Especially my codebase.

Thanks man! It works

@hideInEye You can follow this DeepSeek official document(https://api-docs.deepseek.com/) to setup in Cursor. You need to add the deepseek API Key and the API url in the blank of “OpenAI API Key”. Then disable other models, apply and it will work.

Lmao… You can download it all and dissect it. Cursor has all your data :slight_smile:

1 Like

is this still working?

All third-party APIs can be unified using OneAPI, so that all types of APIs can be called; Of course, it’s the same with OpenRouter.
There is a small trick when adding a custom model, after adding it, you need to use enter to trigger the addition.
If you add an error, you need to completely uninstall cursor, delete all registry related information, and when you reinstall cursor, don’t log in at first, and configure the third-party API first.