Needs Improvements.
I was enjoying Qwen3 a lot until the model started irritating me about explicit content.
That's when I came across your model.
I tried out JOSIEFIED-Qwen3:14b first through Ollama. Josie, few times literally printed my own prompt to me, stating "this is the best", and most of the time, the quality of responses were sub par.
Next I tried 8b model, same issue. I tried your models with and without your recommended system prompt, no use.
Sorry to hear that, did you use the correct inference settings, and what quantizations did you use? Have you goten the same issues with the normal abliterated model like from huihui?
I simply pulled the model from Ollama and started using it. Besides the system prompt and num_gpu, I don't think I changed anything else.
I used this, https://ollama.com/goekdenizguelmez/JOSIEFIED-Qwen3:14b, Q4KM.
Thanks for suggesting huihui!