Getting started

Install and run OpenLLM Studio on your machine.

OpenLLM Studio is a free, MIT-licensed, open-source local LLM app built with Tauri and Rust. It is a GPU-accelerated local LLM app that auto-detects your hardware, recommends the right model and quantization, and downloads GGUF models from Hugging Face through a llama.cpp GUI.

Download OpenLLM Studio View source on GitHub

Download a pre-built binary

The fastest way to start is to download the latest release for your platform. The installer is about 5 MB; runtime binaries are downloaded on demand based on your hardware.

Windows

OpenLLM-Studio_x64-setup.exe

macOS

Apple Silicon & Intel builds

Linux

x86_64 AppImage / deb

On first launch the app detects your GPU, RAM, and CPU, then recommends the best runtime. You only download the inference binary you need.

Build from source

If you prefer to build locally, you will need Node.js 20.19+ (or 22.12+) and the latest stable Rust toolchain.

git clone https://github.com/Icecubesaad/OpenLLM-Studio.git
cd OpenLLM-Studio
npm install
npm run tauri dev

For production builds, run npm run tauri build and look for the bundle in src-tauri/target/release/bundle/.

Install system dependencies

Tauri requires native system libraries. Install the ones for your OS before running or building the app.

Windows

Install the Microsoft Visual C++ Build Tools with the Windows SDK and C++ workload.

macOS

xcode-select --install

Linux (Ubuntu / Debian)

sudo apt update
sudo apt install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf libgtk-3-dev libayatana-appindicator3-dev

Linux (Arch / Fedora)

Install the equivalent Tauri dependencies for your distribution: webkit2gtk-4.1, libappindicator, gtk3, librsvg, and patchelf.

Understand runtime binaries

OpenLLM Studio uses llama.cpp binaries for inference. They are stored in your local app data directory and downloaded only when needed.

Variant	Platform	Use case
CPU	All	No GPU available
CUDA 12.4 / 13.1	Windows, Linux	NVIDIA GPUs
Vulkan	Windows, Linux	AMD / Intel GPUs
Metal	macOS	Apple Silicon

Binary storage paths:

Windows: %APPDATA%\com.openllm.studio\llama-binaries\
macOS: ~/Library/Application Support/com.openllm.studio/llama-binaries/
Linux: ~/.config/com.openllm.studio/llama-binaries/

Add a Hugging Face token

You can browse and download public GGUF models without an account. A token is only needed for private models or higher rate limits.

Create or copy a token from your Hugging Face account.
Open OpenLLM Studio and go to Settings.
Paste the token in the Hugging Face section and save.

Troubleshooting

GPU not detected

Update to the latest GPU drivers. NVIDIA users should install the CUDA Toolkit; AMD users should install the latest drivers from AMD. Then check Settings → General → Hardware Info.

Binary download fails

Check your internet connection, try a different runtime engine in Settings → General, or switch to the CPU variant as a fallback. Firewalls and proxies can also block GitHub releases.

Build errors

Make sure Node.js is 20.19+ or 22.12+, then clear dependencies and the Rust cache:

rm -rf node_modules && npm install
cd src-tauri && cargo clean

Ready to run your first local model?

Download OpenLLM Studio, let the hardware wizard pick a model, and start chatting offline.

Download OpenLLM Studio Read the blog