Feedback
I haven't tried Ataraxy since v1 came out but decided to give this new version a try since some time has passed and damn... It's pretty damn amazing I've gotta say, I would even go as far as to say it is SOTA for its size, it just has no right to be this good when it's so small. You can comfortably rope it to 16k without losing coherence, the prose is superb while at the same time, it's capable of following complex and elaborate instructions. Simply amazing.
I like v4-Advanced, too. It is noticeably better than v1 in my opinion. I find it interesting to chat with, its point of view feels philosophical and deconstructing in its nature. Fairly interesting model, even if a bit too verbose at times. I also think that this model is rather impressive for its size.
I often like its choices of words, too (I encourage it to write more creatively in llama-cli (part of llama.cpp) via --dry-multiplier 0.8 --xtc-probability 0.1)