I just wanted to write this super quick post because I did a search on Perplexity and it didn't give me good results, so here it is.
Groq is the new ultra fast LLM serving plataform, using their LPU chips. I wanted to use it in Cursor because I've been feeling GPT-4 is now terrible for code (sadly, because I already paid the yearly subscription to Cursor mainly to use GPT-4 to code assistance).
To setup Groq you can go direclty to their website and create your account. Then go to Groq Cloud and create your API Key.
And thats all the setup you need to do on Groq, now you can do to Cursor detailed settings and activate the "Using key" option from OpenAI (its compatible with it). And then change the OpenAI Base URL for:
https://api.groq.com/openai/v1
If you have any problems activating your Key, try desactivating all models and add a new model. I'll use Mixtral because its the best model there (and its super cheap). But you can also try Gemma or Lamma (7B/70B) for now.
And thats it, you're good to go. Good coding.
The next thing to try probably is using Claude 3 Haiku, which is also blazing fast and super smart for code.
