https://eithanasulin.github.io/DuckLLM/
- Windows
- macOS
- Linux
- Android
DuckLLM is a free, locally-run AI 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.
- Download DuckLLM Installer.exe
- Agree To The License And Install!
Press
EscTo Close It.
- Download DuckLLM.deb
- Open DuckLLM.deb And Install!
Press
EscTo Close It.
- Download DuckLLM Installer.dmg
- Open DuckLLM Installer.dmg
- Drag DuckLLM Into The Applications Folder
Press
EscTo Close It.
- Privacy First: Fully local execution - no data is sent to external servers.
- Ultra Quick: With DuckLLM's App You'll Experience Highly Impressive Speeds With Well Made Features.
- Smooth UI: With DuckLLM's Dynamic Island-like UI You Experience High Responsiveness & Smooth Animations
Mobile installation currently uses Wllama & Ollama for on-device inference.
- Download DuckLLM from the Google Play Store.
- Open the app, complete or skip the username setup, and navigate to Download Center.
- Choose one of the available models:
- DuckLLM Light (0.6B) - Recommended for most devices
- DuckLLM Base (1.6B)
- DuckLLM Pro (3.1B)
- Hardware Requirements Can Be Found In Requirements.md.
- Documentation Can Be Found In Documentation.md.
Due To Apple's Aggressive Memory Management on Heavy Workflows DuckLLM Might Be Closed By MacOS, To Prevent This DuckLLM Uses a 3.1b Model Only To Manage Memory Better.
For inquiries regarding commercial licensing, please reach out via:
- Email: duckinc68@gmail.com
- Discord: https://discord.com/invite/DkNt6FXf7J (This Only Applys For DuckLLM's Model The App Is Free To-Use)
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.
To Get DuckLLM's Source Code You Can Run
git clone https://github.com/EithanAsulin/DuckLLM.git- 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.