Mugou-9B-GGUF

GGUF export of MugouLab/Mugou-9B.

Quants exported by this notebook: q4_k_m, Q6_K, Q8_0, BF16.

Notes

The source model is expected to be a merged 16-bit Hugging Face/safetensors repo.
Training-time vision layers were preserved/frozen in the original training notebook.
GGUF multimodal inference depends on your llama.cpp/runtime support for Qwen3.5-VL. If image/video input does not work in your GGUF runtime, use the merged HF repo MugouLab/Mugou-9B with Transformers/vLLM/SGLang.
This converter notebook adds CLIP/Qwen-VL default image_mean / image_std metadata when the bundled llama.cpp mmproj converter encounters processor configs that omit those fields.

Question 2

What license applies to MugouLab/Mugou-9B-GGUF?

Accepted Answer

License: apache-2.0. Verify terms on Hugging Face before commercial use.

Question 3

How do I run MugouLab/Mugou-9B-GGUF locally?

Accepted Answer

Download a GGUF file from this page and load it in guIDE or llama.cpp. Pipeline task: image-text-to-text.

Question 4

How much VRAM or disk space does MugouLab/Mugou-9B-GGUF need?

Accepted Answer

Runs locally from ~879.0 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).

File	Type	Quantization	Size	Link
source_merged_hf.BF16-mmproj.gguf	GGUF	GGUF	879.0 MB	Download
source_merged_hf.BF16.gguf	GGUF	GGUF	16.69 GB	Download
source_merged_hf.Q4_K_M.gguf	GGUF	GGUF	5.24 GB	Download
source_merged_hf.Q6_K.gguf	GGUF	GGUF	6.85 GB	Download
source_merged_hf.Q8_0.gguf	GGUF	GGUF	8.87 GB	Download