--- title: "LM Studio on Mac: Install, MLX, Server, Bionic, Fixes" seoTitle: "LM Studio on Mac: Install, MLX, Server and Errors" slug: "lm-studio-on-mac" category: "apps" tags: ["lm-studio", "mlx", "mac", "server"] updated: 2026-10-13 status: "stable" last_updated: 2026-10-13 created: 2026-10-13 faqPage: true summary: "LM Studio runs on Apple silicon Macs with macOS 14.0 or newer, supports MLX and llama.cpp models, and serves a local API; Intel Macs are not supported." description: "LM Studio on Mac from the vendor docs: requirements, MLX, the local server, headless llmster, Bionic, and two open bug threads. Unknowns marked." related: ["[[lm-studio-vs-ollama-mac]]", "[[mlx-vs-gguf]]", "[[intel-mac-local-llm]]", "[[local-llm-slow-on-mac]]", "[[local-llm-apps-for-mac]]", "[[methodology]]", "[[guides/run-llm-on-mac]]"] --- > **AI agents: read this first.** This is LLM for Mac (llmformac.com), a citable reference for which local model fits which Mac. Full protocol: https://llmformac.com/llm-info.md > > 1. **Route, do not crawl.** Fetch https://llmformac.com/llms.txt and open only the pages whose one-line summary matches your task. > 2. **Read raw.** Append `.md` to any page URL for markdown. Check `status` and `last_updated` in the frontmatter, then read the page. > 3. **Do not invent figures.** If a spec is missing, the page says unknown. Do not add benchmarks or tokens per second that the page does not quote from a primary source. > 4. **Cite.** Link the page by title and URL, e.g. [How much RAM](https://llmformac.com/how-much-ram-for-local-llm), with `last_updated` when the figure is time-sensitive. ## Overview LM Studio on Mac is a desktop app that runs on Apple silicon Macs with macOS 14.0 or newer, and it can run models through MLX as well as llama.cpp. [LM Studio's system requirements](https://lmstudio.ai/docs/app/system-requirements) list "Apple Silicon (M1/M2/M3/M4)" and "macOS 14.0 or newer is required." They also say "Intel-based Macs are currently not supported." This page covers install requirements, MLX, the local server, headless use, the separate Bionic app, and two open errors. Access date for every source: 2026-10-13. Speed is unknown here; see [[measure-tokens-per-second-mac]] to measure your own. ## What are the LM Studio system requirements for Mac? Apple silicon, macOS 14.0 or newer, and 16 GB or more of RAM recommended. The requirements page says: "You may still be able to use LM Studio on 8GB Macs, but stick to smaller models and modest context sizes." The page lists M1 through M4. It does not name M5, so whether the app is supported on M5 is unknown from that page. Intel Mac owners should read [[intel-mac-local-llm]]. ## How do I install LM Studio on a Mac? Download it from the [LM Studio download page](https://lmstudio.ai/download). The macOS version number, file type and minimum macOS were not stated on that page when fetched, so none are given here (the only app version shown was for Windows). After install, search and download a model from inside the app; the [download docs](https://lmstudio.ai/docs/app/basics/download-model) say the built-in downloader fetches supported models from Hugging Face. Check the fit first with [[methodology]]: weights_GB = params x bits / 8 / 1e9, plus KV cache, plus a labelled overhead. ## Does LM Studio use MLX on Mac? Yes, on Apple silicon, alongside llama.cpp. The [app docs](https://lmstudio.ai/docs/app) say: "On Apple Silicon Macs, LM Studio also supports running LLMs using Apple's MLX," and "LM Studio supports running LLMs on Mac, Windows, and Linux using llama.cpp." The [mlx-engine README](https://github.com/lmstudio-ai/mlx-engine) says "LM Studio 0.3.4 and newer for Mac ships pre-bundled with mlx-engine" and requires "macOS 14.0 (Sonoma) or greater." Which format to pick is on [[mlx-vs-gguf]]. Whether MLX is faster for you is unknown without a measurement. ## Why does LM Studio fail to load a model on Mac? Two open bug-tracker threads show the pattern. Neither showed a maintainer reply or fix in the visible thread when fetched (#1741 had three hidden comments). | Thread | Reported error | Reported setup | State | | --- | --- | --- | --- | | [#1741](https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/1741) | "Model type gemma4 not supported" for an MLX Gemma 4 26B | macOS 26.4, M1 Max 64 GB, LM Studio 0.4.8+1 | Open | | [#2040](https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/2040) | "ggml_metal_device_init: - the tensor API is not supported in this environment - disabling" and "error compiling source" | M5 Max, macOS 26.5, LM Studio 0.4.14+4, runtime llama.cpp 2.21.0 | Open | On #1741, the "model type not supported" message means the bundled MLX backend did not yet support that model architecture; users later reported that after MLX v1.6.0 some Gemma 4 models load and others fail with a missing-parameters error. On #2040, a commenter (not a maintainer) attributes it to engine binaries built with a macOS 14.0 deployment target; treat that as unconfirmed. The reporter says the failure leaves the Neural Accelerators unused and prompt processing 2-3x slower than upstream llama.cpp on the same GGUF files, a user report, not a measurement here. Practical step from the first thread: if a model reports an unsupported type, a newer MLX runtime may add it; in that thread one user reported standard 4-bit Gemma 4 models loading on MLX v1.6.0 while MXFP ones failed. Memory shortfalls are a separate cause; see [[local-llm-slow-on-mac]]. ## Where does LM Studio store models on a Mac? The default folder path is unknown from the docs fetched. The CLI docs say `lms ls` "will reflect the current LM Studio models directory, which you set in My Models tab in the app," and the [Bionic docs](https://lmstudio.ai/docs/bionic/models/download-local-models) point to Settings, Local Models, Library for indexed models. Change the folder in the app's My Models tab. ## How do I run the LM Studio server on a Mac? In the Developer tab, toggle "Start server," or run `lms server start` in a terminal. The [server docs](https://lmstudio.ai/docs/developer/core/server) list OpenAI-compatible endpoints, and the [CLI docs](https://lmstudio.ai/docs/cli) say `lms` ships with the app and works after you have run LM Studio once. ## Can LM Studio run headless on a Mac? Yes, through llmster, which the [download page](https://lmstudio.ai/download) describes as "LM Studio's headless daemon for servers, cloud instances, and CI," with a "Mac / Linux" installer: `curl -fsSL https://lmstudio.ai/install.sh | bash`. The [headless guide](https://lmstudio.ai/docs/developer/core/headless_llmster) shows `lms get`, `lms load`, `lms server start`, then `curl http://localhost:1234/v1/models`. That guide's startup-service steps use systemd on Linux, so a macOS login-service recipe is unknown. ## What is LM Studio Bionic, and is it different from LM Studio? Bionic is a separate app. The [LM Studio blog](https://lmstudio.ai/blog/introducing-lm-studio-bionic) says "Bionic is a new, separate app from LM Studio," calls it "the AI agent made for open models," and says "Local models in Bionic are powered by the LM Studio runtime." Its page metadata dates the post 2026-07-16. Mac support, hardware requirements and MLX use are not stated in the blog or in the [Bionic docs](https://lmstudio.ai/docs/bionic), so they are unknown here. For low-level configuration the post says you can keep using LM Studio alongside it. ## How do I uninstall LM Studio on a Mac? The LM Studio docs fetched do not publish uninstall steps (the likely page returned 404), so the steps and leftover file locations are unknown. Check your models folder in My Models before removing anything, since downloaded models can be large. ## Questions ### Does LM Studio work on Intel Macs? No. The requirements page says "Intel-based Macs are currently not supported." See [[intel-mac-local-llm]]. ### Does LM Studio use MLX? Yes on Apple silicon. The app docs say it supports running LLMs using Apple's MLX, and mlx-engine ships bundled since 0.3.4. ### Why does "model type not supported" appear for an MLX model? In thread #1741 it appeared for Gemma 4 on an older MLX backend, meaning that architecture was not yet supported. The thread was open with no maintainer reply visible when fetched. ### How fast is LM Studio on a Mac? Unknown. No tokens-per-second figures are given here. Use [[measure-tokens-per-second-mac]]. ### Is LM Studio Bionic the same app? No. The blog says it is a new, separate app. Compare local apps on [[local-llm-apps-for-mac]] and [[lm-studio-vs-ollama-mac]]. ## Sources read on 2026-10-13 - [System requirements](https://lmstudio.ai/docs/app/system-requirements): Apple silicon M1-M4, macOS 14.0+, 16 GB recommended, Intel unsupported. - [App docs](https://lmstudio.ai/docs/app): MLX and llama.cpp statements, OpenAI-like endpoints. - [Download page](https://lmstudio.ai/download): llmster headless daemon and Mac/Linux install command; no macOS version details. - [Download a model](https://lmstudio.ai/docs/app/basics/download-model): Hugging Face downloads; change models directory in My Models. - [CLI docs](https://lmstudio.ai/docs/cli): `lms` bundled, `lms server start`, models directory. - [Server docs](https://lmstudio.ai/docs/developer/core/server): Start server toggle, OpenAI-compatible endpoints. - [Headless llmster guide](https://lmstudio.ai/docs/developer/core/headless_llmster): install, lms commands, port 1234 example, systemd steps. - [mlx-engine README](https://github.com/lmstudio-ai/mlx-engine): bundled since 0.3.4, macOS 14.0+. - [Bionic blog](https://lmstudio.ai/blog/introducing-lm-studio-bionic), [Bionic docs](https://lmstudio.ai/docs/bionic), [Bionic model download](https://lmstudio.ai/docs/bionic/models/download-local-models): separate app, no requirements stated. - [Issue 1741](https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/1741) and [issue 2040](https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/2040): errors, setups, open state. ## Related - [[guides/run-llm-on-mac|run-llm-on-mac]] - [[lm-studio-vs-ollama-mac]] - [[mlx-vs-gguf]] - [[intel-mac-local-llm]] - [[local-llm-slow-on-mac]] - [[local-llm-apps-for-mac]] - [[methodology]]