The repository describes the files as a direct conversion of the original model and lists the base architecture as Wan with 14B parameters.
The page lists quantizations from Q2_K to Q8_0. File sizes range from 5.3 GB for Q2_K to 15.4 GB for Q8_0; intermediate options include Q3, Q4, Q5 and Q6 variants. QuantStack says the files can be used through city96’s ComfyUI-GGUF custom node, placed in ComfyUI/models/unet. The repository is labelled Apache-2.0, while stating that the original model’s licensing terms and usage restrictions still apply.
Hugging Face reports 831,282 downloads in the previous month. The page says no inference provider has deployed the model. It does not establish generation quality, speed, hardware requirements, equivalence with the original weights, or the date on which each quantization was released.