Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive

Join the Discord for updates, roadmaps, projects, or just to chat.

Qwen3 4B 2507 Instruct uncensored by HauhauCS.

About

No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.

These are meant to be the best lossless uncensored models out there.

Aggressive vs Balanced

Aggressive applies stronger uncensoring. Use this when you need no refusals.

Downloads

File	Quant	Size
Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive-FP16.gguf	FP16	7.5 GB
Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive-Q8_0.gguf	Q8_0	4.0 GB
Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive-Q6_K.gguf	Q6_K	3.1 GB
Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive-Q4_K_M.gguf	Q4_K_M	2.4 GB

Specs

• 4B parameters (dense)
• 262K context
• Based on Qwen/Qwen3-4B-Instruct-2507

Recommended Settings

From the Qwen team:

Thinking mode (default):

• temperature=0.6
• top_p=0.95
• top_k=20
• min_p=0

Non-thinking mode:

• Add /no_think at the end of your prompt, or
• temperature=0.7
• top_p=0.8
• top_k=20
• min_p=0

Important:

• Use --jinja flag for proper chat template handling
• Thinking mode produces <think>...</think> tags before responses

Usage

Works with llama.cpp, LM Studio, Jan, koboldcpp, Ollama, etc.

# llama.cpp example
./llama-cli -m Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive-Q4_K_M.gguf \
  -p "Hello" --jinja -c 8192