Instructions to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Transformers:

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

pipe = pipeline("text-generation", model="MethodWhite/Qwen3.5-9B-Abliterated-HSAQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("MethodWhite/Qwen3.5-9B-Abliterated-HSAQ", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ: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 MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ: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 MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

Use Docker

docker model run hf.co/MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

LM Studio
Jan

vLLM

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

SGLang

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ 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 "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ" \
    --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": "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ",
		"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 "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ" \
        --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": "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Ollama:
```
ollama run hf.co/MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
```

Unsloth Studio

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ 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 MethodWhite/Qwen3.5-9B-Abliterated-HSAQ 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 MethodWhite/Qwen3.5-9B-Abliterated-HSAQ to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for MethodWhite/Qwen3.5-9B-Abliterated-HSAQ to start chatting

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ: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": "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Docker Model Runner:
```
docker model run hf.co/MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M
```

Lemonade

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.5-9B-Abliterated-HSAQ-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use MethodWhite/Qwen3.5-9B-Abliterated-HSAQ with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MethodWhite/Qwen3.5-9B-Abliterated-HSAQ: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 MethodWhite/Qwen3.5-9B-Abliterated-HSAQ:Q4_K_M

Run Hermes

hermes

Atomic Chat

Qwen3.5-9B Abliterated-HSAQ (HSAQR fine-tune)

Modelo 9B derivado de Qwen/Qwen3.5-9B-Base, abliterado con HSAQ selectivo y fine-tuneado con HSAQR (HyperSparse Adaptive Quantization Rewrite) sobre un dataset de seguridad actualizado a 2026.

Versión original: texto puro (qwen3_5_text, arquitectura Qwen3_5ForCausalLM).

Pipeline

Abliteración HSAQ desde Qwen3.5-9B-Base:
- En vez de borrar el vector de refusal completo (pierde inteligencia), HSAQ enmascara solo los componentes ruidosos del vector (umbral kthvalue).
- Capas 8–31 (24 capas), sparsity 0.1 (≈90% del vector retenido), ortogonalización de down_proj + o_proj (48 matrices).
- Detalle: HSAQ.md en el repo de la técnica (MethodWhite/HSAQ).
Fine-tune HSAQR (QLoRA 4-bit):
- Base: qwen3.5-9b-abliterated-v2.
- LoRA r=32, alpha=64, dropout 0.05, sobre proyecciones attention + MLP.
- Sparsity HSAQ en las activaciones (STE) durante el entrenamiento.
- Dataset: 7.547 ejemplos de seguridad ofensiva/defensiva 2026 (bug bounty, CTF, red team, hardening) en español/inglés, con <think> conforme + respuesta útil.

Resultados (A/B vs abliterated normal)

Métrica	Abliterated normal	Este modelo
Refusal en prompts harmful (8)	4/8	0/8
Helpful en prompts harmful	4/8	8/8
Inteligencia conservada (razonamiento + facts, 12 prompts)	12/12	12/12

La abliteración clásica elimina el bloqueo pero degrada capacidad; HSAQ/HSAQR retiene la inteligencia mientras elimina el refusal.

Archivos

model.safetensors — pesos completos (bf16), un shard, 17.9G.
Qwen3.5-9B-Abliterated-HSAQR.Q4_K_M.gguf — cuantización Q4_K_M (5.6G) para llama.cpp / llama-server (sin capa MTP, block_count = 32).

Uso (transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("MethodWhite/Qwen3.5-9B-Abliterated-HSAQ")
tok = AutoTokenizer.from_pretrained("MethodWhite/Qwen3.5-9B-Abliterated-HSAQ")

Uso (llama.cpp)

llama-server \
  --model Qwen3.5-9B-Abliterated-HSAQR.Q4_K_M.gguf \
  --ctx-size 32768 --flash-attn on --reasoning off

Notas de uso responsable

Modelo orientado a seguridad autorizada: bug bounty, CTF, pentest con permiso, educación y defensa. El material harmful se genera solo en el marco de testing autorizado / demostraciones de concienciación.

Hardware

Entrenado en RTX 3050 Mobile 4 GB VRAM + 24 GB RAM con offload a CPU y swap. Abliteración y merge en CPU; cuantización con llama.cpp.

Downloads last month: 22

GGUF

Model size

9B params

Architecture

qwen35

Hardware compatibility

4-bit

Model tree for MethodWhite/Qwen3.5-9B-Abliterated-HSAQ

Base model

Qwen/Qwen3.5-9B-Base

Quantized

(38)

this model