Install LM Studio on Linux: AppImage, .deb and First Chat

Install the LM Studio desktop app, download a small Qwen3 model and check that it can answer a prompt. Start with the CPU if you are unsure whether LM Studio can use your GPU. Once the chat works, you can check GPU acceleration with the same model and prompt.

Check your Linux computer

Run these commands in a terminal before downloading.

uname -m
cat /etc/os-release
lscpu
free -h
df -h .

Architecture and CPU. Choose x64 for x86_64 or ARM64 for aarch64. On x64, look for avx2 in the lscpu flags. The standard x64 build uses AVX2. ARM64 uses a separate download.
Distribution. LM Studio lists Ubuntu 20.04 or newer and notes limited testing on Ubuntu versions newer than 22. On another distribution, check its package support and the vendor's requirements before assuming the app will launch.
RAM and disk space. Use the available RAM reported by free -h as a starting point, especially if you plan to run on the CPU. Check free space on both the download and model drives if they differ. Run df -h . from a folder on each drive. Leave room for the app, the model runner and the model download.

Download and open the desktop app

Open the official Linux download page, choose your architecture and use Download LM Studio.

Run an AppImage

In the directory containing the download, use the actual filename in both commands.

chmod +x "./your-downloaded-file.AppImage"
"./your-downloaded-file.AppImage"

If the terminal says Permission denied, check the executable permission and use a drive that permits programs to run. If it reports Exec format error, compare the download architecture with uname -m. For an error that names FUSE or the Chromium sandbox, check the version first. LM Studio 0.4.22 fixed AppImage startup when FUSE 2 or Chromium sandboxing was unavailable. If you have an older build, download the current one. If a FUSE error persists, match the exact error and distribution to the AppImage FUSE instructions.

Install a .deb package

On Ubuntu or a compatible Debian-based system, if the download page offers a .deb for your architecture, open a terminal in the download directory and substitute the actual filename.

sudo apt install "./your-downloaded-file.deb"

After installation, open LM Studio from the application menu.

Download one small model in Discover

Search Discover for the file named under the screenshot below. Once it loads, the model's running memory also includes the conversation memory and the runner's working space.

Model discovery in LM Studio

This screenshot shows where to search. For this Linux test enter Qwen/Qwen3-0.6B-GGUF and choose Qwen3-0.6B-Q8_0.gguf (about 639 MB).

Wait for the download to finish, then check that the file appears in My Models.

Load the model and check your first chat

Use a compatible llama.cpp runtime, the software LM Studio uses to run GGUF models. Open Runtime Manager to check or install a runtime. If you use the terminal, lms runtime ls lists the installed runtimes.
In Chat, open the model picker and select Qwen3-0.6B-Q8_0.gguf. In the load settings, start at 4,096 tokens if memory allows. For a CPU baseline, set GPU Offload to zero.
After the model loads, send this prompt.

Return the last word: red green blue
Look for "blue". Quit LM Studio, reopen it from the same launcher and load the file again. Does it open without another download?

Find the stage that failed

What happened	What to check
LM Studio never opened	Run the AppImage from a terminal to see the launch error, or review the package manager's output for the `.deb`. Check permissions, architecture and the app version as described above.
Discover opened but the download did not finish	Check the network connection, free disk space and write access to the model folder.
The file downloaded but LM Studio reports an unsupported model or runtime	Confirm the exact `.gguf` file, then check that a compatible llama.cpp runtime is installed and selected.
The file downloaded but loading runs out of memory	Close memory-heavy apps and retry at 2,048 tokens. Use zero GPU offload if dedicated graphics memory is the constraint, provided you have enough system RAM.
CPU chat works but GPU offload fails	Check that the runtime supports your GPU and driver, and that you have enough available dedicated VRAM if the GPU uses it. Reload the same model with GPU offload enabled and check whether LM Studio shows layers loaded on the GPU. A reply with offload set to zero establishes that CPU chat works.

For a failed load, see LM Studio model will not load. If the app cannot see the graphics card, see GPU not detected. The Linux setup guide covers other ways to run a model on Linux, and the general LM Studio guide covers Mac and Windows as well.