The repository maintains an uncensored, quantized version of the Qwen‑Image‑2.1 image‑generation model, providing multiple precision formats for local inference via ComfyUI. The conversion uses a stable‑diffusion.cpp commit (1330cebae8f2ba99249df846cc0c9444fcbd4308) and the comfy‑kitchen tooling, delivering GGUF, Safetensors, and Apple‑Silicon‑optimized MLX files. File sizes range from 4.00 GB (MLX 4‑bit) up to 14.23 GB (BF16 GGUF), with the Q4_K_M GGUF (4.60 GB) highlighted as a balanced choice. The accompanying Qwen3‑VL 8B text encoder occupies 17.53 GB in BF16 or 9.35 GB in INT8 ConvRot, while the VAE is 676 MB; the README recommends keeping the diffusion model in GPU VRAM and offloading the INT8 encoder to system RAM to save roughly 9–17 GB of VRAM. The model lacks any built‑in safety filter, making it suitable for adult or sensitive content generation where filtering would otherwise block prompts. Installation requires placing the diffusion, encoder, and VAE files in the appropriate ComfyUI model directories, using the leejet/ComfyUI‑GGUF custom node and CLIPLoader with type “qwen_image”. The project reports 2,188,658 downloads but provides no benchmarked resolution, inference speed, or quality metrics, leaving users to evaluate trade‑offs across quantization levels empirically. The stated license is the Qwen Research License, with commercial‑use terms left undefined.