Instructions to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="XythicK/Mag-Mell-R1-Uncensored-21B-GGUF")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("XythicK/Mag-Mell-R1-Uncensored-21B-GGUF", dtype="auto")

llama-cpp-python

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="XythicK/Mag-Mell-R1-Uncensored-21B-GGUF",
	filename="mag-mell-r1-uncensored-21b-q2_k.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

SGLang

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XythicK/Mag-Mell-R1-Uncensored-21B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with Ollama:
```
ollama run hf.co/XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
```

Unsloth Studio new

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for XythicK/Mag-Mell-R1-Uncensored-21B-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for XythicK/Mag-Mell-R1-Uncensored-21B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for XythicK/Mag-Mell-R1-Uncensored-21B-GGUF to start chatting

Docker Model Runner
How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with Docker Model Runner:
```
docker model run hf.co/XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M
```

Lemonade

How to use XythicK/Mag-Mell-R1-Uncensored-21B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull XythicK/Mag-Mell-R1-Uncensored-21B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Mag-Mell-R1-Uncensored-21B-GGUF-Q4_K_M

List all available models

lemonade list

Mag-Mell-R1-Uncensored-21B-GGUF

🧠 Model Overview

Mag-Mell-R1-Uncensored-21B-GGUF is a quantized version of Mag-Mell-R1-Uncensored-21B, optimized for efficient inference with reduced memory usage and faster runtime while preserving as much of the original model quality as possible.

This repository provides multiple quantized variants suitable for:

Local inference
Low-VRAM GPUs
CPU-only environments

🔗 Original Model

Base model: Mag-Mell-R1-Uncensored-21B
Original repository: https://huggingface.co/JustOnion/Mag-Mell-R1-Uncensored-21B
Original authors: JustOnion

📦 Quantization Details

Quantization method: GGUF
Quantization tool: llama.cpp
Precision: Mixed (2-8,bit depands in variant)
Activation aware: No (weight-only quantinization)
Group size: 256 (K-quant variants)

📦 Available Quantized Files

Quant Format	File Name	Approx. Size	VRAM / RAM Needed	Notes
Q2_K	`mag-mell-r1-uncensored-21b-q2_k.gguf`	~7.8 GB	~8 GB	Extreme compression; noticeable quality loss
Q3_K_S	`mag-mell-r1-uncensored-21b-q3_k_s.gguf`	~9 GB	~10 GB	Smaller, faster, lower quality
Q3_K_M	`mag-mell-r1-uncensored-21b-q3_k_m.gguf`	~10 GB	~11 GB	Better balance than Q3_K_S
Q3_K_L	`mag-mell-r1-uncensored-21b-q3_k_l.gguf`	~10.8 GB	~11.5 GB	Highest-quality 3-bit variant
Q4_0	`mag-mell-r1-uncensored-21b-q4_0.gguf`	~11.7 GB	~12.9 GB	Legacy format; simpler quantization
Q4_K_S	`mag-mell-r1-uncensored-21b-q4_k_s.gguf`	~11.7 GB	~13 GB	Smaller grouped 4-bit
Q4_K_M	`mag-mell-r1-uncensored-21b-q4_k_m.gguf`	~12.4 GB	~14 GB	Recommended default
Q5_0	`mag-mell-r1-uncensored-21b-q5_0.gguf`	~14.1 GB	~16 GB	Higher quality, larger size
Q5_K_S	`mag-mell-r1-uncensored-21b-q5_k_s.gguf`	~14 GB	~15.1 GB	Efficient high-quality variant
Q5_K_M	`mag-mell-r1-uncensored-21b-q5_K_M.gguf`	~14.5 GB	~16 GB	Near-FP16 quality
Q6_K	`mag-mell-r1-uncensored-21b-q6_k.gguf`	~16.8 GB	~18 GB	Minimal quantization loss
Q8_0	`mag-mell-r1-uncensored-21b-q8_0.gguf`	~21.6 GB	~23 GB	Maximum quality; large memory

💡 Recommendation: Start with Q4_K_M for the best quality-to-performance ratio.

🚀 Usage Example

llama.cpp

./main -m mag-mell-r1-uncensored-21b-q5_0.gguf -p "The World is beautiful isn't it?" -n 256

Python (llama-cpp-python)

from llama_cpp import Llama

llm = Llama(
    model_path="<MODEL_FILE>.gguf",
    n_ctx=4096,
    n_threads=8
)

print(llm("Your prompt here"))

🙋 Contact

Maintainer: M Mashhudur Rahim [XythicK]

Role:
Independent Machine Learning Researcher & Model Infrastructure Maintainer

(Focused on model quantization, optimization, and efficient deployment)

For issues, improvement requests, or additional quantization formats, please use the Hugging Face Discussions or Issues tab.

❤️ Acknowledgements

Thanks to the original model authors for their ongoing contributions to open AI research, and to Hugging Face and the open-source machine learning community for providing the tools and platforms that make efficient model sharing and deployment possible.

Downloads last month: 156

GGUF

Model size

20B params

Architecture

llama

Hardware compatibility

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Model tree for XythicK/Mag-Mell-R1-Uncensored-21B-GGUF

Base model

JustOnion/Mag-Mell-R1-Uncensored-21B

Quantized

(3)

this model