should try this one.
first L3 model to keep my interest. passes long-term roleplaying at 6.5k context. behaves like a decent 70b model. can handle complex character cards. doesnt go on weird tangents.
Ill probably be using this as my default model for a time.
I recommend trying it out at Q8 if you can.
Makes it doubly unfortunate that llama.cpp crashes when making imatrix quants of it.
Also, at some point you should consider making a page with your collected reviews :) I know I am guided by such comments when trying out models.
@mradermacher i actually have been looking into starting a page/blog of sorts. sadly im not too tech savvy with website related things especially since im wanting to self host it. but ill figure it out eventually
Best of luck to you, I'll be looking forward to see it :)