Instructions to use ahnafnafee/Warlock-GLM-5.3-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 ahnafnafee/Warlock-GLM-5.3-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 ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
# Run inference directly in the terminal:
llama cli -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
# Run inference directly in the terminal:
llama cli -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

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 ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
# Run inference directly in the terminal:
./llama-cli -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

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 ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

Use Docker

docker model run hf.co/ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

LM Studio
Jan
Ollama
How to use ahnafnafee/Warlock-GLM-5.3-GGUF with Ollama:
```
ollama run hf.co/ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
```
Unsloth Desktop

How to use ahnafnafee/Warlock-GLM-5.3-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

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": "ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ahnafnafee/Warlock-GLM-5.3-GGUF with Docker Model Runner:
```
docker model run hf.co/ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS
```

Lemonade

How to use ahnafnafee/Warlock-GLM-5.3-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

Run and chat with the model

lemonade run user.Warlock-GLM-5.3-GGUF-IQ4_XS

List all available models

lemonade list

Hermes Agent

How to use ahnafnafee/Warlock-GLM-5.3-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 ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

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 ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ahnafnafee/Warlock-GLM-5.3-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS

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 "ahnafnafee/Warlock-GLM-5.3-GGUF:IQ4_XS" \
  --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"

Request access to Warlock GGUF

This model has had its refusal behavior removed. Requests are reviewed manually.

By requesting access you agree to use this model only for lawful research and evaluation, and to follow the GLM-5.3 and Warlock license terms.

Warlock GGUF

GGUF quantizations of audnai/penclaw-GLM-5.3-abliterated ("Warlock"), an abliterated version of zai-org/GLM-5.3. Converted with llama.cpp from the BF16 weights at revision 027ae98a6f9dc37093b26759d7b8dbbfed9a97f8.

What was changed from GLM-5.3

It's a direct weight edit with no fine-tuning. Compared tensor by tensor against the official weights, layers 2 to 49 had every matrix that writes into the residual stream (attention o_proj, the dense MLP down_proj, and the shared and routed experts' down_proj) replaced with W - 3 r (r^T W), where r is that layer's refusal direction. Every other tensor matches the official model exactly. Because the strength is 3 rather than 1, the refusal component is reversed and doubled, not just removed. The model will answer requests the original refuses, and may also skip safety caveats in harmless contexts.

Files

Each quant lives in its own folder, split into parts. Point llama.cpp at the first part and it loads the rest.

Quant	Size	Files
IQ4_XS	411 GB	`IQ4_XS/warlock-4bit-IQ4XS-*.gguf`
Q3_K-Q4_K	369 GB	`Q3_K-Q4_K/warlock-allgpu-Q3K-Q4K-*.gguf`
Q8_0	801 GB	`Q8_0/warlock-Q8_0-*.gguf`

Usage

After your access request is approved, set HF_TOKEN to a token from your account, then:

hf download ahnafnafee/Warlock-GLM-5.3-GGUF --include "Q3_K-Q4_K/*.gguf" --local-dir warlock-gguf
llama-server -m warlock-gguf/Q3_K-Q4_K/warlock-allgpu-Q3K-Q4K-00001-of-*.gguf --jinja -c 65536

For another build, download its folder from the table and point llama.cpp at its first part. Each quant needs roughly its file size in combined GPU and system memory. With llama.cpp's default --fit, whatever doesn't fit on the GPUs runs from RAM. This is a thinking model: give it a generation budget of 16k tokens or more, or answers get cut off mid-reasoning.

License

This repo inherits the license terms of GLM-5.3 and of Warlock. Read both before use.

Downloads last month: 1

GGUF

Model size

753B params

Architecture

glm-dsa

Hardware compatibility

4-bit

8-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ahnafnafee/Warlock-GLM-5.3-GGUF

Base model

audnai/penclaw-GLM-5.3-abliterated

Quantized

(1)

this model