Plenty of better and better models coming out all the time. Right now I recommend, depending on what you can run:
7B: Openhermes 2 Mistral 7B
13B: XWin MLewd 0.2 13B
XWin 0.2 70B is supposedly even better than ChatGPT 4. I’m a little skeptical (I think the devs specifically trained the model on gpt-4 responses) but it’s amazing it’s even up for debate.
Plenty of better and better models coming out all the time. Right now I recommend, depending on what you can run:
7B: Openhermes 2 Mistral 7B
13B: XWin MLewd 0.2 13B
XWin 0.2 70B is supposedly even better than ChatGPT 4. I’m a little skeptical (I think the devs specifically trained the model on gpt-4 responses) but it’s amazing it’s even up for debate.