LM Studio download: choose a build for your platform
Five installer variants cover every supported platform. Pick the one that matches your hardware, verify the download, install, and you will be running a local model within minutes.
Recap Capsule
The LM Studio download decision comes down to three questions: what operating system, what processor architecture, and do you need a portable build? The table below maps every combination to the correct installer. Read the platform notes below the table if you are unsure about any of those three factors.
Choosing the right LM Studio download
One wrong choice — Intel build on an Apple Silicon Mac, for example — still installs and runs, but misses the GPU acceleration that makes LM Studio usable at full speed on that hardware.
LM Studio ships separate builds for different hardware targets, not just different operating systems. This matters most on macOS, where Apple Silicon (M-series) and Intel chips require different binaries to get native GPU acceleration. Downloading the wrong macOS variant means running through Rosetta 2 translation, which works but is slower and does not use Metal GPU acceleration. On Windows, the distinction is simpler: there is one installer for 64-bit Windows, which covers the vast majority of machines made in the last decade. On Linux, the AppImage format abstracts over most distribution differences, but there is a FUSE dependency to be aware of on minimal installs.
The installer itself is relatively small — typically 150 to 350 MB depending on platform and version. Model files are not bundled in the installer; they download separately through the in-app library after the application is running. A typical 7B model in Q4_K_M quantization adds 4 to 4.5 GB, so make sure the drive where LM Studio stores models has at least 5 to 10 GB free before your first session.
Platform download guide
Five download targets cover all supported platforms: Windows x64, macOS Apple Silicon, macOS Intel, Linux AppImage, and the Windows portable build for restricted environments.
| Platform | File type | Approx. installer size | First-run note |
|---|---|---|---|
| Windows 10 / 11 (x64) | .exe installer | ~200–280 MB | Run the .exe, follow the wizard; CUDA detected automatically if NVIDIA driver is current |
| macOS Apple Silicon (M1–M4) | .dmg disk image | ~200–320 MB | Drag to Applications; Metal GPU acceleration is automatic — no driver install needed |
| macOS Intel (x86_64) | .dmg disk image | ~200–300 MB | Same install flow as Apple Silicon; GPU acceleration limited to iGPU on most Intel Macs |
| Linux (AppImage) | .AppImage | ~150–250 MB | chmod +x the file before launching; install libfuse2 if FUSE is missing |
| Windows portable build | .zip archive | ~220–290 MB | Extract and run — no installer, no registry writes; useful for external drives or restricted machines |
Windows download and install
The Windows LM Studio download is a standard .exe installer that handles all setup steps automatically, including GPU backend detection for NVIDIA cards.
On Windows 10 or 11, run the downloaded .exe file. The installer wizard asks for an install location and creates a Start menu entry. It does not require administrator privileges by default on most Windows configurations, though UAC may prompt depending on your system policy. After the wizard completes, launch LM Studio from the Start menu.
CUDA GPU acceleration requires an NVIDIA driver that supports the CUDA version bundled with the LM Studio release. If the GPU does not appear in the LM Studio status bar after launch, update the NVIDIA driver to the latest version available from NVIDIA's website, restart the machine, and relaunch LM Studio. AMD GPU support on Windows uses Vulkan as the primary path; ROCm on Windows has limited support compared to Linux. See the Windows platform page for complete hardware compatibility notes.
macOS download and install
The macOS LM Studio download comes as a .dmg — drag to Applications, launch, and Metal GPU acceleration activates automatically on both Apple Silicon and Intel chips that support it.
On macOS, open the downloaded .dmg file, drag the LM Studio application icon to the Applications folder, and eject the disk image. Launch LM Studio from Applications or Spotlight. On macOS 13 and later, Gatekeeper may show a security prompt on first launch — click Open in the prompt to proceed. No additional configuration is required for GPU acceleration on Apple Silicon; the Metal backend is active by default.
The key question on macOS is which build to download. Check your chip under the Apple menu → About This Mac. If it says M1, M2, M3, M4 (or any variant), download the Apple Silicon build. If it says Intel Core i5, i7, i9, or similar, download the Intel build. The Apple Silicon build does not run natively on Intel Macs, and the Intel build runs through Rosetta 2 on Apple Silicon rather than using Metal. See the Mac platform page for system requirements detail.
Linux download and install
The Linux LM Studio download is an AppImage — a single portable file that runs on most distributions without a package manager, provided FUSE is available.
Save the .AppImage file to a convenient location. Open a terminal, navigate to that directory, and run:
chmod +x LMStudio-*.AppImage
./LMStudio-*.AppImageIf the launch fails with a FUSE-related error, install the required library. On Ubuntu and Debian-based systems: sudo apt install libfuse2. On Fedora: sudo dnf install fuse. No reboot is needed after installing FUSE. For GPU acceleration on Linux, NVIDIA CUDA and AMD ROCm both require their respective runtime libraries to be installed separately — see the Linux platform page for the specific package names and version requirements.
Verifying your download
Each LM Studio release ships with SHA-256 checksums for every platform installer. Verifying the checksum before running the installer confirms the file is complete and unmodified.
The checksum file is available alongside each platform installer on the upstream GitHub releases page. After downloading both the installer and the checksum file, run the verification command for your platform. On macOS or Linux: shasum -a 256 LMStudio-*.AppImage (substitute the actual filename). On Windows: CertUtil -hashfile LMStudio-*.exe SHA256. Compare the output to the expected hash in the checksum file — if they match, the download is complete and unmodified.
For guidance on software supply chain security and verification practices, the NIST software verification guidelines and the FTC's AI transparency guidance both address the responsible handling of downloaded software in professional and consumer contexts.
After your LM Studio download: recommended next steps
The installer gets the application running; the first model download and a quick test prompt confirm that inference is working correctly on your hardware.
Once LM Studio is installed and open, the recommended sequence is: open the Discover tab, search for a 7B instruct model, download the Q4_K_M variant, load it, and send a short test prompt. The quickstart guide walks through this in five steps and typically takes under ten minutes. If you want a more structured walkthrough that also covers server mode, use the tutorial instead.
If the model fails to load or the GPU is not detected, the troubleshooting page covers the five most common post-install problems with concrete fixes. The platform-specific pages for Windows, Mac, and Linux include hardware compatibility notes and driver requirements for each OS.
Frequently asked questions
Five questions from readers who are preparing to download LM Studio for the first time.
If your Mac has an Apple Silicon chip (M1, M2, M3, or M4 series), download the Apple Silicon build — it uses Metal GPU acceleration and runs significantly faster than the Intel build on those chips. If your Mac has an Intel processor, download the Intel (x86_64) build. Check Apple menu → About This Mac to see which chip your machine has.
The LM Studio installer is typically 150 to 350 MB depending on platform and version. This is the application only — model files download separately through the in-app library. A 7B model in Q4_K_M quantization adds approximately 4 to 4.5 GB when downloaded inside the app, so ensure you have at least 5 to 10 GB of free storage before your first session.
Each release includes SHA-256 checksums for the platform-specific installers, available on the upstream GitHub releases page. After downloading, verify with shasum -a 256 on macOS or Linux, or CertUtil -hashfile on Windows. A matching checksum confirms the file is complete and unmodified.
Open the Discover tab and download a 7B instruct model in Q4_K_M quantization — this is the recommended first model for most hardware. Then follow the quickstart guide to load the model and send a first prompt. The full tutorial adds server-mode setup as a next step for developers who want to connect LM Studio to external tools.
Yes. A portable build for Windows runs from a folder without writing to the system registry or requiring administrator privileges. Extract the archive and run the executable directly. This is useful for running LM Studio from an external drive or in environments where installing software requires authorization you do not have.