Qt6 / QML native
HiDPI, Wayland, system theming. Feels at home on GNOME and KDE.
Out now · Ubuntu Linux
An offline LLM app for Ubuntu Linux.
Available now on Ubuntu — native Qt6, Vulkan-accelerated. Gemma 4 E4B (Q4) on your GPU. No cloud, no account.
Ubuntu installation
Install Localyze in three steps.
- Open the Ubuntu download and save the
.debfile. - Double-click the downloaded file. Ubuntu Software opens automatically.
- Click Install, enter your password, then launch Localyze from Applications.
First launch takes a little longer.
The app downloads the Gemma 4 E4B model, roughly 4 GB, once. Keep the app open on Wi-Fi until it completes. After that, chat works offline and no Terminal setup is required.
Built for Linux
Linux-native, GPU-accelerated.
Qt6 + QML on the desktop. Vulkan on the GPU. A proper .deb installer for Ubuntu.
Vulkan GPU
AMD, Intel, NVIDIA — all accelerated through one open API.
Distro-friendly
Ubuntu users get a normal .deb installer that opens in Ubuntu Software.
Native tech
Built with the Linux stack.
- Qt6 + QML — native widgets, Wayland support, HiDPI, dark mode.
- Vulkan — vendor-neutral GPU inference on AMD, Intel, and NVIDIA.
- Ubuntu .deb — double-click the download and install through Ubuntu Software.
Other platforms: Local AI for Windows · Local AI for Mac
Offline LLM on Linux, answered.
Is there an AI app for Linux that runs offline?
Yes. Localyze.ai is a native Qt6 / QML app that runs Gemma 4 E4B (Q4) on-device, no cloud needed.
Which distros are supported?
Ubuntu 22.04+ on x86_64 is officially supported. The packaged .deb opens in Ubuntu Software.
How is inference accelerated?
Vulkan on the GPU across AMD, Intel, and NVIDIA. CPU fallback is available but slow.
How much does the Linux app cost?
Free to download. No subscriptions, no tiers, no account, no caps. One model download, then fully offline.