onurburak9/Qwen3.8-27B-Uncensored-MLX ? reverse-engineered prompt

Build me a clean model repo for the Qwen3.8 27B uncensored MLX build.

I want the repository to clearly present the different quantized versions in separate folders for 2 bit, 4 bit, 6 bit, and 8 bit, and have a polished README that explains what the model is, what it’s for, and how to use it on Apple Silicon with mlx vlm. Please include simple download and run examples for text and image input, plus a quick note that the 4 bit version is the best default and the 2 bit one is mostly just for archival use.

Make sure the page reads nicely for people who just want to try the model, and that it includes the important safety note that this is meant for research and red teaming, not general safe deployment. If anything is missing from the docs or metadata, fill it in in a straightforward way and look up current docs online if you need to.