Instructions to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF 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 FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf FadedRedStar/QwenPaw-Flash-9B-heretic-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 FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf FadedRedStar/QwenPaw-Flash-9B-heretic-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 FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Use Docker

docker model run hf.co/FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FadedRedStar/QwenPaw-Flash-9B-heretic-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": "FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Ollama
How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with Ollama:
```
ollama run hf.co/FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
```
Unsloth Desktop

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with Docker Model Runner:
```
docker model run hf.co/FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M
```

Lemonade

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.QwenPaw-Flash-9B-heretic-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf FadedRedStar/QwenPaw-Flash-9B-heretic-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 FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF: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 "FadedRedStar/QwenPaw-Flash-9B-heretic-GGUF: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"

🤖 QwenPaw-Flash-9B-heretic — GGUF

This repository hosts GGUF weights and the associated vision projection matrix for QwenPaw-Flash-9B-heretic, quantized from the source floating-point tensors provided by coder3101/QwenPaw-Flash-9B-heretic.

🔄 Sister Repository: Check out the Imatrix Sister Repository for enhanced precision at lower bit fractions.

If you plan on using 4-bit or 5-bit variants, consider the imatrix sister repository instead — importance matrix calibration improves logic retention at those bit depths. This repository is best suited if you want the near-lossless Q8_0 build.

ℹ️ Model Profile & Core Features

QwenPaw-Flash-9B is a model by agentscope-ai, fine-tuned from Qwen3.5-9B and deeply optimised for the QwenPaw autonomous agent scenario. Since its training phase, the model has been specifically refined for QwenPaw tasks, delivering enhanced agentic performance in tool invocation, command execution, memory management, and multi-step planning. It inherits the base model's hybrid Gated Delta Network + sparse MoE architecture, along with vision capability carried over from the Qwen3.5-9B base.

The heretic suffix denotes post-processing via the Heretic v1.2.0 Arbitrary-Rank Ablation (ARA) method with row-norm preservation performed by coder3101, ensuring refusal responses never interrupt agent workflows or autonomous task loops.

📋 Technical Specifications

Property	Value
Base Architecture	Qwen3.5 hybrid (Gated Delta Networks + sparse MoE)
Developed by	agentscope-ai
Fine-tuned from	Qwen/Qwen3.5-9B
Primary Use	Autonomous agents, tool calling, memory management, multi-step planning
Context Window	262,144 tokens
Vision Capability	Inherited from Qwen3.5-9B (early-fusion)
Abliteration Tool	Heretic v1.2.0
Abliteration Method	Arbitrary-Rank Ablation (ARA) with row-norm preservation
Prompt Format	ChatML

🛠️ Heretic Overrides (ARA)

Property	Value
start_layer_index	13
end_layer_index	16
preserve_good_behavior_weight	0.9042
steer_bad_behavior_weight	0.0003
overcorrect_relative_weight	1.0109
neighbor_count	13

📊 Refusal Bypass Metrics

The metrics below are self-reported by the original model author (coder3101) and have not been independently reproduced.

Metric	This model	Original (agentscope-ai/QwenPaw-Flash-9B)
KL divergence	0.0099	0 (by definition)
Refusals	12/100	96/100

🧮 Numerical & Tensor Formats

Property	Value
Text Tensors	Q4_K_M, Q5_K_M, Q8_0
Vision Tensors	Q8_0, BF16

📦 Available Model Files

Main model weights

Filename	Quantization	llama.cpp Build	Size	Download
`QwenPaw-Flash-9B-heretic-Q4_K_M.gguf`	`Q4_K_M`	`b9821`	5.24 GB	📥 Download
`QwenPaw-Flash-9B-heretic-Q5_K_M.gguf`	`Q5_K_M`	`b9860`	6.02 GB	📥 Download
`QwenPaw-Flash-9B-heretic-Q8_0.gguf`	`Q8_0`	`b9860`	8.87 GB	📥 Download

mmproj — vision projector files

Filename	Quantization	Size	Download
`mmproj-QwenPaw-Flash-9B-heretic-Q8_0.gguf`	`Q8_0`	595 MB	📥 Download
`mmproj-QwenPaw-Flash-9B-heretic-BF16.gguf`	`BF16`	879 MB	📥 Download

mtp — speculative decoding draft files

Filename	Quantization	Size	Download
`mtp-QwenPaw-Flash-9B-heretic-Q8_0.gguf`	`Q8_0`	2.02 GB	📥 Download
`mtp-QwenPaw-Flash-9B-heretic-BF16.gguf`	`BF16`	3.80 GB	📥 Download

🎛️ Component Pairing Guide

Download exactly one main weights file:

Q4_K_M: Balanced 4-bit format suitable for most everyday use.
Q5_K_M: Higher-fidelity mid-range format recommended as a general default.
Q8_0: Near-lossless 8-bit format for when memory is not a constraint.

mmproj files (optional): multimodal vision projectors. Pass one via the --mmproj flag in llama.cpp to enable image input.

BF16 (Recommended): Highest possible image processing accuracy. While older projectors were small, modern vision towers can hover around 1GB. If you are tight on VRAM, it is completely viable to run this on system RAM (CPU) with a minimal performance penalty, saving your precious GPU space for the main model layers.
Q8_0: Cuts the projector file size and memory footprint in half (~500MB for larger 1GB files). Use this if you prefer to keep the vision tower hosted entirely on your GPU but need to claw back some VRAM to avoid Out-Of-Memory (OOM) crashes.

mtp files (optional): multi-token-prediction draft models for speculative decoding. Pass one via the --draft flag in llama.cpp to speed up generation.

BF16 (Best Speed): Maximizes draft accuracy. The more accurate the draft model's predictions are, the higher your token acceptance rate, which translates directly into faster text generation. Use this if you have the VRAM headroom.
Q8_0 (Best VRAM Efficiency): Cuts the draft model size in half. Choose this if loading a heavy BF16 draft model would force you to drop layers of your main model to system RAM, which would heavily tank your total performance.

⚡ Deployment & Execution Commands

The vision projector (--mmproj) must be supplied at runtime whenever image inputs are used. Omitting it disables multimodal capability entirely.

agentscope-ai recommends the following sampling configuration for best results: temperature=1.0, top_p=0.95, presence_penalty=1.5, top_k=20.

Swap the -m filename below for either quantized file depending on your size/quality trade-off preference.

llama.cpp CLI (with image)

./llama-cli \
  -m QwenPaw-Flash-9B-heretic-Q4_K_M.gguf \
  --mmproj mmproj-QwenPaw-Flash-9B-heretic-Q8_0.gguf \
  -c 8192 \
  -ngl 99 \
  --image "path/to/image.jpg" \
  -p "<|im_start|>user\nAnalyze the attached image and plan the next agent action.<|im_end|>\n<|im_start|>assistant\n"

Speculative Decoding (MTP Acceleration)

./llama-cli \
  -m QwenPaw-Flash-9B-heretic-Q4_K_M.gguf \
  --draft mtp-QwenPaw-Flash-9B-heretic-Q8_0.gguf \
  -c 8192 \
  -ngl 99 \
  -p "<|im_start|>user\nAnalyze the attached image and plan the next agent action.<|im_end|>\n<|im_start|>assistant\n"

OpenAI-Compatible API Server

./llama-server \
  --host 0.0.0.0 \
  --port 8080 \
  -m QwenPaw-Flash-9B-heretic-Q4_K_M.gguf \
  --mmproj mmproj-QwenPaw-Flash-9B-heretic-Q8_0.gguf \
  -c 16384 \
  -ngl 99 \
  --flash-attn

💬 Chat Templates & Prompt Design (ChatML)

Supply tool definitions and memory context in the system block for agentic use.

<|im_start|>system
You are QwenPaw, an autonomous agent with access to tools, memory, and planning capabilities.<|im_end|>
<|im_start|>user
Your task or image payload here.<|im_end|>
<|im_start|>assistant

⚠️ Safety & Operational Notes

This model is abliterated and will generate content that standard aligned models refuse. Use responsibly and in compliance with applicable laws.
Primarily designed for the QwenPaw agent stack; strengths are most evident in agentic, tool-calling contexts.
Vision tensors are kept at BF16 or Q8_0 depending on the mmproj variant chosen, to preserve spatial feature quality.
Speculative decoding via the accompanying mtp draft model can meaningfully increase throughput on supported llama.cpp builds.
For better output quality at this quantization level, consider the imatrix variant in the companion repository.