REVIEW · LOCAL AI APP
Verdict: LM Studio is one of the easiest ways to run AI models on your own Mac, and it can work entirely offline once you have the model files. Its Chat with Documents feature reads a .docx, .pdf or .txt file you attach; it does not search your Mac for a document you cannot find.
27 October 2026 · By Ishtiaque Hossain · 4 minute read
Looking for a file you can describe but not name? Try Shlexball. The first 30 searches are free and need no account.
macOS 15 or later. Apple silicon. 16 GB memory. On a phone? Email yourself the link.
This review is based on LM Studio's own pages, read on 14 October 2026. Every claim links to the page it came from. We are the makers of Shlexball, so we say plainly where LM Studio does better.
| What it is | An app for running local AI models, now with the Bionic agent for work and code |
| Runs on | Apple silicon Macs (M1 to M4), macOS 14.0 or newer; Intel Macs not supported |
| Memory | 16 GB or more recommended; 8 GB Macs should stick to smaller models |
| Documents | Attach .docx, .pdf or .txt files to a chat |
| Offline | Can run entirely offline once you have downloaded model files |
| Best for | Trying open AI models privately on your own hardware |
LM Studio is an app for downloading and running large language models (LLMs, the kind of AI behind chatbots) on your own computer. Its site names open models such as Qwen, Mistral, Gemma and gpt-oss.
The current release presents Bionic, an agent for work and code: it creates and edits documents with every change saved automatically, and helps with coding tasks, automations and computer control.
Under the hood it uses the LM Studio runtime with MLX and llama.cpp, two engines for running models efficiently. Bionic also offers real-time voice transcription, and LM Studio says that voice and audio are processed locally and never leave your device.
You can attach document files (.docx, .pdf and .txt) to a chat in LM Studio. That gives the model extra context, so you can ask questions about what the documents say.
If a document is short enough to fit in the model's context, its working memory, LM Studio adds the whole file to the conversation. Context is measured in tokens, and one token is often about 3/4 of a word.
If a document is too long, LM Studio uses Retrieval-Augmented Generation, or RAG: it tries to fish out the relevant parts of a long document, or several documents, and hands those to the model.
LM Studio is candid that RAG sometimes works really well and sometimes needs tuning. Its tip is to put as much context in your question as you can, including terms and ideas you expect to appear in the document.
On a Mac, LM Studio needs Apple silicon (M1, M2, M3 or M4) and macOS 14.0 or newer. Intel-based Macs are not supported.
LM Studio recommends 16 GB of memory or more. You may still be able to use it on an 8 GB Mac, but you should stick to smaller models and modest context sizes.
For the most demanding tasks, Bionic can run large open models such as GLM 5.2, Kimi K3 and DeepSeek V4 Pro, which need far more power than everyday chat models.
Running locally is the point. LM Studio says it can operate entirely offline, as long as you download some model files first, and that privacy is key to its ethos.
It also runs cloud services, which it says are zero data retention across the board, meaning requests are not kept. The pages we read do not list prices, so check the site before you plan around a cost.
Only the ones you give it. Chat with Documents works on files you attach to a chat, and RAG searches inside those attached files for the passages that matter.
That makes it good for asking questions about a contract or a report you already have open, and the wrong tool for finding which file on your Mac holds that contract in the first place.
Shlexball is the step before: describe the file you remember in plain English and it finds it on your Mac, searching file and folder names and the text inside files, including PDFs and Word documents.
Like LM Studio, it runs a language model on your Mac and never uploads files. It needs an Apple silicon Mac, macOS 15 or later and 16 GB of memory, and the first 30 searches are free.
Shlexball searches your whole Mac and hands you the file: 30 free searches.
▶ DOWNLOAD SHLEXBALLYes. Attach .docx, .pdf or .txt files to a chat; short documents are read in full and long ones are searched with RAG for the relevant parts.
Yes. LM Studio can operate entirely offline once you have downloaded model files.
No. On the Mac it needs Apple silicon (M1 to M4) and macOS 14.0 or newer; Intel-based Macs are not supported.
LM Studio recommends 16 GB or more. On 8 GB Macs it may still work if you use smaller models and modest context sizes.