LM Studio for Mac: macOS install for Apple Silicon and Intel
A complete walkthrough for getting lm studio mac up and running — from .dmg download and Gatekeeper to Metal acceleration, model storage paths, and a full uninstall.
Capsule Summary
LM Studio mac requires macOS 13 or later. Apple Silicon Macs benefit from Metal GPU acceleration across the unified memory pool — M1 through M4 chips all work well. Install by dragging the app from the .dmg to Applications, bypass Gatekeeper once in System Settings, and models land in ~/.cache/lm-studio/models unless you change the path.
Installing LM Studio from the .dmg
The macOS installation is a standard drag-and-drop .dmg workflow — no package manager, no terminal commands required.
The lm studio mac build ships as a .dmg disk image. Download the appropriate build from the download page — there are separate builds for Apple Silicon (arm64) and Intel (x86_64). Downloading the wrong build still works on macOS through Rosetta 2 translation, but native performance is significantly better, so matching the build to your chip is worth the extra second of attention.
Open the downloaded .dmg file. A Finder window appears with the LM Studio application icon and an alias to the Applications folder. Drag LM Studio onto the Applications alias. The copy takes a few seconds. Eject the .dmg, then open LM Studio from Applications or Spotlight. On first launch, macOS will verify the app signature — this takes a moment but only happens once.
To confirm which chip your Mac has, open the Apple menu and select "About This Mac." The chip field will show M1, M2, M3, M4, or an Intel Core designation. Download the matching installer variant from the download page to avoid any Rosetta overhead.
Apple Silicon optimisation and Metal acceleration
The unified memory architecture on M-series chips removes the GPU VRAM ceiling that limits discrete GPU machines, letting large models fit comfortably in shared RAM.
On Apple Silicon, LM Studio uses Apple's Metal API for GPU-accelerated inference. Metal on M-series Macs is fast — a 7B model at Q4 quantization on an M2 MacBook Air produces token rates that rival a midrange NVIDIA GPU on Windows, without a fan spinning. The unified memory architecture is the key advantage: the neural engine and GPU cores draw from the same physical memory pool as the CPU, so a 16 GB M2 machine can run a 7B model with comfortable headroom while the OS and browser remain responsive.
A 32 GB M2 or M3 machine opens up 13B models at full quality. A 64 GB M3 Max or M4 Max pushes into the 30B range at usable token speeds. These are not theoretical limits — they reflect real-world usage patterns with Q4_K_M and Q5_K_M quantizations. LM Studio's layer slider is less relevant on Apple Silicon because the unified memory means there is no separate VRAM pool to spill out of, but it is still available for fine-tuning the balance between GPU and neural engine layers.
Intel Macs do not have Metal GPU acceleration at the same level. On Intel builds, LM Studio falls back to CPU inference using AVX2 and AVX-512 instruction sets where available. Token rates are slower than Apple Silicon, but the app is fully functional and suitable for lighter models (7B at Q4 or smaller).
Gatekeeper and first-launch permissions
Gatekeeper blocks unrecognized publishers by default; the fix is a one-time click in System Settings rather than disabling Gatekeeper entirely.
On first launch, macOS Gatekeeper may display a message saying LM Studio "cannot be opened because it is from an unidentified developer" or similar. This happens when a newly notarized certificate has not yet propagated to Apple's reputation database, or when your Mac's security settings require explicit approval for apps from outside the App Store.
The resolution: open System Settings (macOS 13+) or System Preferences (older macOS), navigate to Privacy & Security, scroll down to the Security section, and click "Open Anyway" next to the LM Studio entry. A confirmation dialog appears; click Open. From that point on, LM Studio launches normally with no further prompts. You do not need to disable Gatekeeper globally — the per-app exception is sufficient.
If the "Open Anyway" button does not appear, the app may have been quarantined. In Terminal, run xattr -d com.apple.quarantine /Applications/LM\ Studio.app to remove the quarantine flag, then try launching again. The MIT AI Lab at MIT publishes useful background on code signing and trust chains (mit.edu) for readers who want the underlying security rationale.
Model storage in ~/.cache on macOS
Models are stored as ordinary files in a predictable cache path, making backup, migration, and manual management straightforward.
By default, LM Studio stores all downloaded model files in ~/.cache/lm-studio/models. This is a hidden folder in your home directory. To navigate there in Finder, press Cmd+Shift+G and type the path, or use Go › Go to Folder from the menu bar. Each model family gets its own subfolder, and the GGUF weight files sit directly inside.
If your startup disk is limited, you can redirect the model directory to an external drive or a secondary volume. Open Settings inside LM Studio, find the "Model directory" field, and paste in your preferred path. The app respects symlinks, so creating a symlink from the default cache path to an external drive also works without touching the setting.
Settings and chat presets live in ~/Library/Application Support/LM Studio. This folder is separate from the model cache, so clearing the cache does not reset your configuration.
Uninstalling LM Studio from macOS
A full uninstall touches three locations — the app bundle, the model cache, and the settings folder.
Drag LM Studio from /Applications to the Trash to remove the application bundle. This does not remove model files or settings. To complete the cleanup, also delete ~/.cache/lm-studio and ~/Library/Application Support/LM Studio. Model files can be large — confirm the folder sizes in Finder before deleting.
Tools like AppCleaner (a third-party utility) can find all files associated with an application automatically, including support files that live in less obvious locations. Running AppCleaner against LM Studio after dragging it to Trash is a reliable way to ensure nothing is left behind.
| Mac model | Chip | Recommended model size |
|---|---|---|
| MacBook Air / MacBook Pro 2020–2021 | M1 (8–16 GB) | 7B Q4; 13B Q4 on 16 GB |
| MacBook Air / MacBook Pro 2022–2023 | M2 (8–96 GB) | 7B–13B on 16 GB; 30B on 32 GB; 70B Q4 on 96 GB |
| MacBook Pro / Mac Studio 2023 | M3 (8–128 GB) | 13B on 16 GB; 30B on 32 GB; 70B on 64 GB+ |
| MacBook Pro / Mac Studio / Mac Pro 2024–2025 | M4 (16–192 GB) | 30B comfortable on 32 GB; 70B on 64 GB; 100B+ on max configs |
| Intel MacBook Pro / iMac | Intel Core (8–64 GB) | 7B Q4 on 16 GB; 13B Q4 on 32 GB (CPU-only, slower throughput) |
Frequently asked questions
Common questions about installing and using LM Studio on macOS across Apple Silicon and Intel hardware.
LM Studio for Mac requires macOS 13 Ventura or later. Both Apple Silicon (M1, M2, M3, M4) and Intel builds are available from the download page. macOS 14 Sonoma and macOS 15 Sequoia are fully supported on both chip families.
On Apple Silicon Macs, LM Studio uses the Metal API and Apple's unified memory architecture. Because CPU and GPU share the same memory pool, large models fit in addressable RAM and benefit from GPU compute at the same time. This typically delivers token rates that rival a midrange discrete GPU on Windows, without the VRAM ceiling.
Open System Settings, go to Privacy & Security, scroll to the Security section, and click "Open Anyway" next to the LM Studio entry. Confirm in the dialog that follows. You only need to do this once; subsequent launches open without prompts. You do not need to disable Gatekeeper globally.
Model files are stored in ~/.cache/lm-studio/models by default. You can change this path in the app's Settings. Each model family gets its own subfolder containing the downloaded GGUF files. Use Cmd+Shift+G in Finder to navigate to hidden paths.
Drag LM Studio from /Applications to the Trash to remove the application. Then manually delete ~/.cache/lm-studio for model files and ~/Library/Application Support/LM Studio for settings. A tool like AppCleaner can locate all associated files automatically to make sure nothing is left behind.