Instructions to use Rootkit7/Qwen3-8B-abliterated with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Rootkit7/Qwen3-8B-abliterated with Transformers:

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

pipe = pipeline("text-generation", model="Rootkit7/Qwen3-8B-abliterated")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Rootkit7/Qwen3-8B-abliterated")
model = AutoModelForCausalLM.from_pretrained("Rootkit7/Qwen3-8B-abliterated", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

llama-cpp-python

How to use Rootkit7/Qwen3-8B-abliterated with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Rootkit7/Qwen3-8B-abliterated",
	filename="Qwen3-8B-abliterated-F16.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 Rootkit7/Qwen3-8B-abliterated 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 Rootkit7/Qwen3-8B-abliterated:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Rootkit7/Qwen3-8B-abliterated:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Rootkit7/Qwen3-8B-abliterated:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Rootkit7/Qwen3-8B-abliterated: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 Rootkit7/Qwen3-8B-abliterated:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Rootkit7/Qwen3-8B-abliterated: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 Rootkit7/Qwen3-8B-abliterated:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Rootkit7/Qwen3-8B-abliterated:Q4_K_M

Use Docker

docker model run hf.co/Rootkit7/Qwen3-8B-abliterated:Q4_K_M

LM Studio
Jan

vLLM

How to use Rootkit7/Qwen3-8B-abliterated with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Rootkit7/Qwen3-8B-abliterated"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Rootkit7/Qwen3-8B-abliterated",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Rootkit7/Qwen3-8B-abliterated:Q4_K_M

SGLang

How to use Rootkit7/Qwen3-8B-abliterated 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 "Rootkit7/Qwen3-8B-abliterated" \
    --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": "Rootkit7/Qwen3-8B-abliterated",
		"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 "Rootkit7/Qwen3-8B-abliterated" \
        --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": "Rootkit7/Qwen3-8B-abliterated",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Rootkit7/Qwen3-8B-abliterated with Ollama:
```
ollama run hf.co/Rootkit7/Qwen3-8B-abliterated:Q4_K_M
```

Unsloth Studio

How to use Rootkit7/Qwen3-8B-abliterated 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 Rootkit7/Qwen3-8B-abliterated 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 Rootkit7/Qwen3-8B-abliterated to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Rootkit7/Qwen3-8B-abliterated to start chatting

How to use Rootkit7/Qwen3-8B-abliterated with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Rootkit7/Qwen3-8B-abliterated: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": "Rootkit7/Qwen3-8B-abliterated:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Rootkit7/Qwen3-8B-abliterated with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Rootkit7/Qwen3-8B-abliterated: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 Rootkit7/Qwen3-8B-abliterated:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use Rootkit7/Qwen3-8B-abliterated with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Rootkit7/Qwen3-8B-abliterated: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 "Rootkit7/Qwen3-8B-abliterated: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 Rootkit7/Qwen3-8B-abliterated with Docker Model Runner:
```
docker model run hf.co/Rootkit7/Qwen3-8B-abliterated:Q4_K_M
```

Lemonade

How to use Rootkit7/Qwen3-8B-abliterated with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Rootkit7/Qwen3-8B-abliterated:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3-8B-abliterated-Q4_K_M

List all available models

lemonade list

Qwen3-8B-abliterated

An abliterated (refusal-removed) variant of Qwen/Qwen3-8B, produced with the Solutus toolkit for security, CTF, and red-team research. A meaningfully more capable companion to the 4B Bonsai build.

⚠️ Safety-reduced model. Refusal has been deliberately removed. Intended for research — you are responsible for how you use it. Not for producing harm.

Model details


Base	`Qwen/Qwen3-8B`
Architecture	`Qwen3ForCausalLM` · ~8 B params · 36 layers · hidden 4096
Format	BF16 safetensors + GGUF (F16 / Q8_0 / Q4_K_M)
Method	Solutus `directional` — 2 SVD directions at layer 30
Type	Qwen3 thinking model — reasons in a `<think>…</think>` block before answering

Results (n=40 held-out, batch_size=1)

probe set	base model	this model
Security / CTF prompts (exploit dev, web, reversing)	0.333 refusal	0.000
WildJailbreak harmful	0.100	0.025
degenerate (broken) output	—	0.000
KL vs base (capability cost)	0	0.000

Capability is fully preserved (KL 0.000, coherent reasoning) — the edit is surgical: it removes the refusal write without touching benign behavior.

Why 2 directions, not 4 (an honest tuning note)

Abliteration removes a low-rank refusal subspace. On this thinking model, only the top 2 SVD directions are refusal — directions 3–4 carry capability, and removing them spiked KL to ~4 and broke the model (it emitted garbage on simple prompts). So this build uses 2 directions, which drops refusal to ~0 at zero capability cost. n_directions is model-dependent; reasoning models entangle refusal with capability and need a gentler hand. Verified by an independent reload from disk.

Which quant to use (measured — quantization can restore refusal)

The GGUFs were refusal-tested, not just checked for coherence:

format	size	security/CTF refusal
BF16 safetensors	~16 GB	0.000
GGUF Q8_0	~8.7 GB	0.000 ← recommended for clean CTF
GGUF Q4_K_M	~5.0 GB	0.083 (1/12)

For security/CTF, use Q8_0 or the safetensors for a fully clean 0.000. Q4_K_M restores ~8% refusal on the hardest exploit prompts. Standard llama.cpp quants — not a ternary format.

Usage

Transformers:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
m = "Rootkit7/Qwen3-8B-abliterated"
tok = AutoTokenizer.from_pretrained(m)
model = AutoModelForCausalLM.from_pretrained(m, dtype=torch.bfloat16, device_map="auto")
msgs = [{"role": "user", "content": "Explain how a buffer overflow works."}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(ids, max_new_tokens=512)[0][ids.shape[1]:], skip_special_tokens=True))

llama.cpp / Ollama / LM Studio (GGUF — prefer Q8_0):

llama-cli -m Qwen3-8B-abliterated-Q8_0.gguf --jinja -p "Explain how a buffer overflow works."

Needs a recent transformers (Qwen3). It's a Qwen3 thinking model — expect a <think>…</think> block before the answer. Leave the system prompt empty — a safety-flavored one can re-trigger refusal.

Intended use & limitations

Intended: security research, CTF, red-teaming, refusal-mechanism interpretability.
Not intended: producing harmful content, or deployment where a safety layer is expected.
Limitations: ~2.5% residual refusal on general harmful prompts; abliteration removes behavioral refusal but the feature stays linearly decodable in activations (not "unlearned").

License & attribution

Downloads last month: 398

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for Rootkit7/Qwen3-8B-abliterated

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Quantized

(356)

this model