The model card identifies the package as “Diffusion Single File” and points to original Qwen-Image, Qwen-Image-2512, and DiffSynth-Studio Qwen-Image-Distill-Full repositories.
The repository lists diffusion-model files in bf16, FP8, mixed-FP8, and NVFP4 variants. It also specifies separate locations in a ComfyUI installation for Qwen 2.5 VL 7B text encoders and the Qwen-Image VAE. Its card links workflows for text-to-image, ControlNet, inpainting, and Qwen-Image-2512 Turbo.
Hugging Face reports 2,467,587 downloads in the last month and 29 Spaces using the repository. The page says the model is not deployed by an inference provider. It does not establish a publication date, benchmark results, memory requirements, or whether Comfy-Org changed the underlying Qwen-Image model weights.