How do I get started with Ollama or LM Studio?
Install one app, pick a file that fits your memory and run your first model.
Short answer
Install one app, download a model file that fits your memory, and run it with one click or one command.
- LM Studio: A desktop app with a built-in model browser.
- Ollama: A command-line tool with a small app; one command downloads and runs a model.
- First model: Qwen3.5 9B in Q4_K_M: a 5.7 GB download, about 15 GB of memory on a 32 GB+ Mac.
Published file size (unsloth/Qwen3.5-9B-GGUF, read Oct 17, 2026)
Ollama or LM Studio
Both run models on your own computer and are built on llama.cpp (Hugging Face · Ollama, read Oct 17, 2026) (Hugging Face · LM Studio, read Oct 17, 2026).
- Ollama
- Command line first, with a small appGood for developers: one command downloads and runs a model, and a local API at localhost:11434 lets other apps use it.Ollama README
- LM Studio
- Desktop app with a model browserGood for a first try: search, pick a file from a list that marks what suits your hardware, and chat. On Macs it also runs MLX files.Hugging Face · LM Studio
1. Install
Ollama on macOS or Linux, from the Terminal (or download the app from ollama.com) (Ollama README, read Oct 17, 2026):
curl -fsSL https://ollama.com/install.sh | shOllama on Windows, from PowerShell:
irm https://ollama.com/install.ps1 | iexLM Studio: download the app from lmstudio.ai. It runs on Macs with Apple silicon, Windows and Linux.
2. Pick a file that fits
Q4 file, short conversations. Planning estimates in whole GB.
The download is not the whole story: the app needs working room on top of the file, and the computer needs memory for itself. Start with a size whose estimate is well below your memory:
| Model size | Q4 weights, est. (GB) | 32 GB+ Mac or PC (GB) | Graphics card (GB) |
|---|---|---|---|
| 4B | 2.5 | 11 | 5 |
| 9B | 5.6 | 15 | 9 |
| 27B | 16.9 | 29 | 23 |
For a first run, Qwen3.5 9B in Q4_K_M asks for about 15 GB on a Mac with 32 GB or more (smaller Macs reserve less) and 9 GB on a graphics card. Qwen3.5 9B
3. Run it
Ollama can run any GGUF repository on Hugging Face by name. Add the file version after a colon; without one it picks Q4_K_M when the repository has it (Hugging Face · Ollama, read Oct 17, 2026):
ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q4_K_MIn LM Studio, search the model name (⌘ + Shift + M on a Mac, Ctrl + Shift + M on a PC), open the file list and download Q4_K_M; then load it and chat. From the Terminal, its lms tool takes the file version after @ (Hugging Face · LM Studio, read Oct 17, 2026):
lms get https://huggingface.co/unsloth/Qwen3.5-9B-GGUF@Q4_K_MIf the model answers slowly or the computer starts to struggle, pick a smaller size or a smaller file before anything else.