Official Homepage

https://eithanasulin.github.io/DuckLLM/

Supported Systems

• Windows
• macOS
• Linux
• Android

What is DuckLLM?

DuckLLM is a free, locally-run LLM designed with a strong focus on privacy and security, without compromising on performance or functionality. It simplifies the process of self-hosting an AI model on your own device - for both desktop and mobile - with a privacy-first architecture that ensures your data never leaves your machine.

Getting Started

For End Users (Standalone)

1. Download: Get the latest release for your platform - DuckLLM.exe (Windows), .dmg (macOS), or .AppImage (Linux).
2. Run the Setup Wizard: The wizard will automatically handle Ollama and Python installation if they are not already present.
3. Select a Model: Choose between Full (7.6B) or Light (3.1B) depending on your hardware.

For Developers (Source Setup)

1. Clone the repository and navigate to the project directory.
2. Install dependencies: npm install
3. Launch the app: npm start - performs background dependency checks and starts immediately.
4. Run the setup wizard manually: npm run setup - useful for maintenance or reinstallation.
5. Build a portable executable: npm run dist - output will be placed in the /dist folder.

Alternative: Python Manual Installer

You can also run the installer directly from source:

python src/installer.py

Key Features

• Privacy First: Fully local execution - no data is sent to external servers.
• Fast Boot: Background dependency checks enable near-instant startup.
• RTL Support: Improved Right-to-Left (Hebrew) text alignment and caret handling, optimized for both Windows and Linux.

DuckLLM Mobile

Mobile installation currently uses Wllama for on-device inference.

1. Download DuckLLM from the Google Play Store.
2. Open the app, complete or skip the username setup, and navigate to Download Center.
3. Choose one of the available models:
   • DuckLLM Light (0.6B) - Recommended for most devices
   • DuckLLM Base (1.6B)
   • DuckLLM Pro (3.1B)

Keyboard Shortcuts (Desktop)

To launch DuckLLM via a keyboard shortcut:

• Linux: Add a custom shortcut in System Settings pointing to: python ~/DuckLLM/DuckLLM.py
• Windows: Use PowerToys or a desktop shortcut pointing to: python ~/Desktop/DuckLLM/DuckLLM.py

Press Del to show the window and your configured key to hide it.

Relationship to Qwen

DuckLLM is built on Qwen 2.5 as its base model and extends it through fine-tuning aimed at improving performance in specific areas where the base model underperforms. Training a large language model from scratch was outside the scope of this project; the additional training is focused on targeted improvements rather than architectural changes.

Commercial Use

For inquiries regarding commercial licensing, please reach out via:

• Email: duckinc68@gmail.com
• Discord: https://discord.com/invite/DkNt6FXf7J

Source Code

DuckLLM's source code is available in the /src folder of this repository and is also included in the official releases.

Contributors

• psale

License

• DuckLLM Proprietary License: Covers the DuckLLM model weights. Free for personal use; a commercial license is required for business use.
• Apache 2.0: Covers the Qwen 2.5 base model.