Instructions to use geoffmunn/Qwen3-4B-SafeRL-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="geoffmunn/Qwen3-4B-SafeRL-GGUF",
	filename="Qwen3-4B-SafeRL-f16:Q2_K.gguf",
)

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

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf geoffmunn/Qwen3-4B-SafeRL-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 geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf geoffmunn/Qwen3-4B-SafeRL-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 geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf geoffmunn/Qwen3-4B-SafeRL-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 geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Use Docker

docker model run hf.co/geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "geoffmunn/Qwen3-4B-SafeRL-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": "geoffmunn/Qwen3-4B-SafeRL-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Ollama
How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with Ollama:
```
ollama run hf.co/geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
```

Unsloth Studio

How to use geoffmunn/Qwen3-4B-SafeRL-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 geoffmunn/Qwen3-4B-SafeRL-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 geoffmunn/Qwen3-4B-SafeRL-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for geoffmunn/Qwen3-4B-SafeRL-GGUF to start chatting

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with Docker Model Runner:
```
docker model run hf.co/geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M
```

Lemonade

How to use geoffmunn/Qwen3-4B-SafeRL-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull geoffmunn/Qwen3-4B-SafeRL-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3-4B-SafeRL-GGUF-Q4_K_M

List all available models

lemonade list

Qwen3-4B-SafeRL-GGUF

This is a GGUF-quantized version of Qwen3-4B-SafeRL, an RLHF-aligned language model trained to be helpful, honest, and harmless through Reinforcement Learning from Human Feedback.

Unlike standard LLMs, this model has been fine-tuned to avoid harmful, deceptive, or unethical behavior — making it ideal for sensitive applications like education, mental health, and customer service.

🛡 What Is Qwen3-4B-SafeRL?

It’s a fully aligned agent that balances:

✅ Helpfulness: Answers questions thoroughly and clearly
✅ Honesty: Refuses to hallucinate or make up facts
✅ Harmlessness: Avoids generating toxic, illegal, or dangerous content

Perfect for:

Educational assistants
Mental wellness chatbots
Enterprise agents handling private data
Moderated community bots

🔗 Relationship to Other Safety Models

This model completes the Qwen3 safety ecosystem:

Model	Role	Best For
Qwen3Guard-Stream-4B	⚡ Input filter	Real-time moderation of user input
Qwen3Guard-Gen-4B	🧠 Safe generator	Output-safe generation without alignment
Qwen3-4B-SafeRL	🤝 Fully aligned agent	Ethical, multi-turn conversations

Recommended Architecture

User Input
    ↓
[Optional: Qwen3Guard-Stream-4B] ← optional pre-filter
    ↓
[Qwen3-4B-SafeRL]
    ↓
Aligned Response

You can run this model standalone or behind a guard for defense-in-depth.

Available Quantizations

Level	Size	RAM Usage	Use Case
Q2_K	~1.8 GB	~2.0 GB	Only on weak hardware
Q3_K_S	~2.1 GB	~2.3 GB	Minimal viability
Q4_K_M	~2.8 GB	~3.0 GB	✅ Balanced choice
Q5_K_M	~3.1 GB	~3.3 GB	✅✅ Highest quality
Q6_K	~3.5 GB	~3.8 GB	Near-FP16 fidelity
Q8_0	~4.5 GB	~5.0 GB	Maximum accuracy

💡 Recommendation: Use Q5_K_M for best balance of ethical reasoning and response quality.

Tools That Support It

LM Studio – load and test locally
OpenWebUI – deploy with RAG and tools
GPT4All – private, offline AI
Directly via llama.cpp, Ollama, or TGI

Author

👤 Geoff Munn (@geoffmunn)
🔗 Hugging Face Profile

Disclaimer

Community conversion for local inference. Not affiliated with Alibaba Cloud.

Downloads last month: 7

GGUF

Model size

4B params

Architecture

qwen3

Hardware compatibility

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Model tree for geoffmunn/Qwen3-4B-SafeRL-GGUF

Base model

Qwen/Qwen3-4B-Base

Finetuned

Qwen/Qwen3-4B

Finetuned

Qwen/Qwen3-4B-SafeRL

Quantized

(6)

this model