LM Studio for Linux: install and run with the AppImage
A practical guide to getting lm studio linux running — from FUSE dependencies and AppImage permissions through ROCm AMD acceleration, Vulkan fallback, and desktop launcher integration.
Reader Takeaways
LM Studio on Linux is distributed as an AppImage — no package manager required. Mark the file executable, satisfy the FUSE 2 dependency (or use --appimage-extract-and-run), and the app runs on Ubuntu, Debian, Fedora, and Arch. ROCm powers AMD GPU acceleration; Vulkan covers everything else. Desktop integration takes one extra step but is optional.
Downloading and running the AppImage
An AppImage bundles all dependencies into a single self-contained file — there is no install step, no root required, and no package manager interaction needed.
The lm studio linux build is distributed as an .AppImage file. Download it from the download page. The file is named something like LMStudio-x.y.z.AppImage. After downloading, mark it executable before trying to run it:
chmod +x LMStudio-x.y.z.AppImage
Then launch it directly:
./LMStudio-x.y.z.AppImage
The first run extracts and caches some internal files, which takes a few extra seconds. Subsequent launches are faster. LM Studio stores models and settings in the same default locations as on other platforms — see the storage section below.
FUSE dependency — what it is and how to install it
FUSE 2 is the one dependency AppImages require; it is absent on some newer distributions by default and needs a one-command fix.
AppImages mount their internal filesystem using FUSE. Most AppImages target FUSE 2 (libfuse2). Ubuntu 22.04 ships with FUSE 3 by default, which means libfuse2 must be installed separately:
sudo apt install libfuse2
Ubuntu 24.04 (Noble) has further tightened sandbox rules and may still require the additional flag:
./LMStudio-x.y.z.AppImage --appimage-extract-and-run
That flag extracts the AppImage contents to a temporary directory and runs from there, bypassing the FUSE mount entirely. It is slightly slower to start and uses more disk space temporarily, but works on any kernel version. On Fedora and Arch, FUSE 2 support is available through the fuse and fuse2 packages respectively.
Distribution-specific notes
Tested configurations vary by distro; the table below summarises what works and where you may need extra steps.
Ubuntu 22.04 LTS is the most thoroughly tested platform. The only extra step is installing libfuse2. Ubuntu 24.04 LTS works with the --appimage-extract-and-run flag. Debian 12 (Bookworm) follows the same pattern as Ubuntu 22.04. Fedora 40 ships with FUSE support out of the box; sudo dnf install fuse fuse-libs covers any gaps. Arch Linux users install fuse2 from the official repositories.
On any distro, if the AppImage fails to launch with an error mentioning libz.so, libstdc++, or another system library, the most reliable fix is the --appimage-extract-and-run flag, which unpacks the bundled libraries instead of relying on the system versions. The W3C's technical architecture group provides relevant background on portable application formats (w3.org) for those interested in the broader ecosystem context.
ROCm setup for AMD GPUs
ROCm unlocks GPU-accelerated inference on supported AMD cards, bringing token speeds that rival NVIDIA CUDA on equivalent hardware.
AMD GPU acceleration on lm studio linux runs through ROCm. First, install the ROCm runtime for your distribution. AMD publishes official packages for Ubuntu LTS and RHEL-compatible systems; for other distros, community-maintained packages are available. Check that your AMD GPU is in the ROCm supported hardware list — RX 6000 and RX 7000 series are generally well-supported as of ROCm 5.7+.
After installing ROCm, ensure your user account is in the video and render groups:
sudo usermod -aG video,render $USER
Log out and back in, then launch LM Studio. Open Settings and confirm that the ROCm backend appears under Inference Backend. If it does not, check that /dev/kfd and /dev/dri are accessible to your user.
Vulkan fallback
Vulkan is a cross-vendor GPU API that LM Studio uses when CUDA and ROCm are unavailable, covering older GPUs and Intel integrated graphics.
When neither CUDA nor ROCm is detected, LM Studio automatically tries the Vulkan backend. Vulkan is supported on a wide range of hardware including older NVIDIA cards, AMD cards without full ROCm support, and Intel integrated and discrete GPUs. Install Vulkan drivers for your distribution — on Ubuntu, sudo apt install vulkan-tools mesa-vulkan-drivers covers most Intel and AMD setups; NVIDIA users need the proprietary driver package.
Vulkan inference is slower than CUDA or ROCm on equivalent hardware, but substantially faster than CPU-only. For 7B models at Q4, Vulkan on a modest GPU can still produce multiple tokens per second, making it viable for everyday use.
Desktop integration
Creating a .desktop launcher makes LM Studio searchable from the application menu and allows setting a custom icon.
Move the AppImage to a stable location first — ~/Applications/LMStudio.AppImage is a common choice. Then create a .desktop file:
~/.local/share/applications/lmstudio.desktop
The file should contain the standard desktop entry fields pointing to the AppImage path, with Exec set to the full path and Icon pointing to an extracted icon. The tool AppImageLauncher automates this entire process — it intercepts the first AppImage launch, offers to integrate it, and creates the .desktop file for you. On GNOME and KDE, the integrated app then appears in the application grid and is searchable by name.
| Distribution | Tested version | Notes |
|---|---|---|
| Ubuntu | 22.04 LTS | Install libfuse2 via apt; CUDA and ROCm fully supported |
| Ubuntu | 24.04 LTS | Use --appimage-extract-and-run flag; otherwise fully functional |
| Debian | 12 (Bookworm) | Same as Ubuntu 22.04; install libfuse2 from apt |
| Fedora | 40 | Install fuse fuse-libs via dnf; ROCm available via AMD repo |
| Arch Linux | Rolling (2024+) | Install fuse2 from official repos; AppImageLauncher available in AUR |
Frequently asked questions
Common questions about running LM Studio on Linux distributions using the AppImage format.
Download the AppImage, mark it executable with chmod +x LMStudio-*.AppImage, then run it directly. Ubuntu 22.04 requires libfuse2 installed via apt. Ubuntu 24.04 uses FUSE 3 by default, so pass --appimage-extract-and-run if you hit a FUSE version mismatch on first launch.
FUSE (Filesystem in Userspace) lets the AppImage mount its internal filesystem without root permissions. Without FUSE 2 (libfuse2), the AppImage cannot extract its own contents and will fail to start. Install libfuse2 on Debian/Ubuntu-based systems, or use the --appimage-extract-and-run flag as a root-free alternative.
Yes, through ROCm. Install the ROCm runtime matching your driver version and add your user to the video and render groups. LM Studio detects the ROCm backend automatically. Supported AMD GPUs include RX 6000 and RX 7000 series cards with recent Linux drivers.
LM Studio falls back to Vulkan for GPU acceleration on cards without CUDA or ROCm support. Vulkan covers a wide range of older NVIDIA, AMD, and Intel integrated graphics. Throughput is lower than CUDA or ROCm but still significantly faster than CPU-only inference for most cards.
Move the AppImage to a stable location such as ~/Applications/, then create a .desktop file in ~/.local/share/applications/ pointing to the AppImage path. The tool AppImageLauncher automates this process, handling the .desktop file and icon extraction so the app appears in your desktop environment's application menu.