Quote:
Originally Posted by Marshal
was there a specific reason why you went exactly with those? trying to learn something here. 
what exact mistral model do you recommend? |
Yes performance and uncensored. For my application mistral works better
1 more natural
2 good data in the 7 billion model
3 great speed
4 obedient it does what you want it to do not its own way of doing what you want it to do.
Llama 3 is the closest to gpt 4 but it is censored. So for porn no good.
edit sorry i forgot to post the mistral model
Nous-Hermes-2-Mistral-7B-DPO.Q4_0.gguf
so far so good with that one.