trohrbaugh

/

Qwen3.5-9B-heretic-v2

like 28

I just got this and it's great.
I've not noticed any significant issues yet and I didn't really run in to any topic which would trigger any refusal.
Regardless, because it's not trained/tuned on certain sensitive topics, it probably won't make great choice in those specific use cases, but for general use it looks to be de-censored enough to completely clear all those annoying safety warnings and disclaimers which og qwen likes to periodically insert at random, so at least so far responses appear to be a whole lot cleaner.

i tried the gguf version, Still not the perfect model, spouts a lot of nonsense when asked to write a 1k short story,
waiting for finetunes with rp and storytelling dataset.