Local & self-hosted appsDevelopment & infrastructureLM Studio
LM Studio
Decision matrix
LM Studio runtime decision matrix
Overview
Verdict: the easiest way for a non-developer to run an AI model on their own laptop, free for work use. Pick it for private drafting and document chat on a 16 GB machine. Pick something else if you need open-source software or a shared team server.
Other good fits
Chatting with documents that must stay on your computer.
Giving other apps a local, OpenAI-compatible endpoint later, without changing tools.
+ Pros
Read pros
LM Studio is a desktop app that finds, downloads and runs open models (Llama, DeepSeek, Qwen, Phi and many more) on your own computer. It shows which models fit your hardware before you download them, gives you a chat window, and lets you attach documents that are processed offline. It uses llama.cpp and, on Apple Silicon, Apple's MLX engine. The same app can act as a local server so that other programs can use the model.
Pick it when
You want to try a local model in ten minutes without a terminal.
You work with text that should not go to a cloud service, such as unpublished results, personal data or HR drafts.
You want to compare two or three open models on your own task before committing to one.
You may later connect a tool (an automation, a script, an editor) to a local model: LM Studio already speaks the OpenAI API format.
− Cons
Read cons
LM Studio is closed source, so you rely on the vendor's terms rather than code you can check. Answer quality is capped by the model your hardware can hold: a laptop model is fine for summarising and rewording, weaker on reasoning and long documents. Intel Macs are not supported. It is a single-user desktop app; there are no user accounts or sharing.
The model should run quietly in the background for scripts and automations: use Ollama.
A team needs one shared assistant with logins: put Open WebUI on a server.
Avoid if
You need one assistant shared by a whole team with logins; run Open WebUI on a server.
You have an Intel Mac; it is not supported.
Requirements
LM Studio supports Apple Silicon macOS, x64/ARM64 Windows and x64 Linux according to its current documentation. Choose a compatible model and runtime with enough memory and storage; no model-specific hardware-fit measurement was made for this review.
First run
Documentation-based suggestion — not run here.
Read first run
Test whether a local model is good enough for your weekly summary. The book A model on your own laptop walks through the idea.
Install LM Studio and download two models it marks as a good fit, for example one 4B and one 8B model.
Take last week's internal update or meeting notes (about one page).
Ask each model: "Summarise in five bullets for a colleague who missed the meeting. Keep names, dates and numbers exactly."
Attach the same page as a document in a new chat and ask three factual questions about it.
Note speed and mistakes for each model.
It worked if one model gets every name, date and number right and answers the three questions from the document. That model is your local default; keep harder tasks for a cloud assistant.
Integration
Applications can use LM Studio's native REST API or its OpenAI-compatible API. Headless llmster is a separate no-GUI route; local serving does not make every model, extension or update path offline.
Check the result
Compare the two items with the original quantities and facts. If a model will not load, select a smaller supported model; one response is not a model-quality result.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on LM Studio's pricing page, app terms, application documentation and system requirements, checked 23 September 2026. The curriculum team has not installed or run LM Studio for this review; the first project is a suggested exercise.
LM Studio pricing · Sep 23, 2026 — Free plan with local models; Bionic+ $20/month and Pro $100/month for hosted open models; team billing via credits.
LM Studio app terms · Sep 23, 2026 — Licensed for personal and internal business use at no charge; proprietary source code.
LM Studio documentation · Sep 23, 2026 — llama.cpp and MLX engines, offline document chat, MCP client, OpenAI-compatible and REST APIs, headless llmster.
System requirements · Sep 23, 2026 — Apple Silicon with macOS 14+, Windows x64/ARM with AVX2, Ubuntu 20.04+; 16 GB RAM and 4 GB VRAM recommended.
Further review notes & alternatives
Costs and operating effort
Prices checked 23 September 2026:
Plan
Price
What you get
Free
$0
The full app with local models, offline voice transcription, limited web search
Bionic+
$20/month
Adds hosted open models and web search
Pro
$100/month
Five times the Bionic+ usage limits
The local app is free for personal and internal business use. You pay for the machine: LM Studio recommends 16 GB of RAM, and on Windows at least 4 GB of dedicated graphics memory. Upkeep is light: accept app updates and delete unused models, which take several gigabytes each. The paid Bionic plans send prompts to US-hosted models; LM Studio says this uses zero data retention, but it is no longer local.
Alternatives and trade-offs
Jan is open source (Apache 2.0) with a similar chat-first design; LM Studio has the better model browser and MLX support.
Ollama is the usual choice for a background service that other tools call; it has a simpler app and optional cloud models.
AnythingLLM is stronger if your main job is chatting with many documents in workspaces.
