Getting Started

Step-by-step guide for first-time local AI users. No special knowledge required — everything explained from scratch.

Complete beginner? Start with basics/ — what AI is, how models work, what hardware you need.

🇷🇺 Russian version: getting-started.ru.md

← Local models · Running models →

Contents

1. What are local models and why use them
2. How to open the terminal
3. Check your Mac
4. Install Homebrew
5. Install Ollama
6. Run your first model
7. Install LM Studio (no terminal alternative)
8. Whats next

1. What are local models and why use them

A local AI model is a program that runs directly on your computer. You download it once, and it works fully offline, for free, with no limits.

Local vs Cloud

What you will be able to do after this guide:

• Ask questions to the model directly in the terminal (like a chat)
• Connect the model to a code editor for autocompletion
• Use the model for translation, text analysis, writing code
• All of this — for free and without internet

2. How to open the terminal

The terminal is a window where you type commands as text instead of clicking buttons. Many commands in this guide need to be entered in the terminal.

Finding Terminal on Mac

Method 1 — Spotlight (fast):

1. Press Cmd (⌘) + Space — search appears
2. Type “Terminal”
3. Press Enter

Method 2 — Finder:

1. Open Finder
2. Go → Utilities
3. Double-click “Terminal”

What the terminal looks like

You will see something like:

MacBook-Air:~ user$

or

user@MacBook-Air ~ %

This is the command prompt. It means the terminal is ready for your command.

Your first command

Type or paste this command into the terminal (right-click → Paste, or Cmd+V):

echo "Hello, Im running an AI model!"

Press Enter. The terminal should respond with:

Hello, Im running an AI model!

Important: Commands starting with $ in guides should be typed without the $. It just represents the prompt. E.g., $ brew install ollama → type brew install ollama.

3. Check your Mac

What chip do you have?

uname -m

• arm64 → Apple Silicon (M1/M2/M3/M4) — great, models run fast
• x86_64 → Intel Mac — still works, slower on large models

How much RAM?

sysctl hw.memsize | awk '{print "RAM: " $2 / 1073741824 " GB"}'

Example output: RAM: 16 GB

Detailed info

system_profiler SPHardwareDataType | grep -E "Chip|Memory|Processor"

Example output:

Chip: Apple M1
Memory: 16 GB

RAM guide

Your RAM	Suitable models
8 GB	Small models (up to 4B params) — Qwen 3.5 4B
16 GB	Mid-size (7B–9B) — universal choice
24–32 GB	Large models (14B–32B)
48+ GB	Very large (70B+)

Most base-model MacBook Air/Pro have 16 GB RAM — optimal for local models.

4. Install Homebrew

Homebrew is the “app store” for the terminal. It installs software not found in the App Store.

Check if already installed

brew --version

• If you see Homebrew 4.x.x → skip to Install Ollama
• If you see command not found: brew → install it with the command below.

Install Homebrew

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

After pressing Enter:

1. Terminal may ask for your Mac password — type it (characters wont show, thats normal). Press Enter.
2. Installation takes 1–5 minutes.
3. Wait for Installation successful!

Verify installation

brew --version

Should show:

Homebrew 4.x.x

Apple Silicon PATH fix

On M1/M2/M3/M4 you might see: Warning: /opt/homebrew/bin is not in your PATH...

echo 'eval "$(/opt/homebrew/bin/brew shellenv)"' >> ~/.zshrc
source ~/.zshrc

Then verify again:

brew --version

5. Install Ollama

Ollama is the simplest program for running AI models. It can:

• Download models on demand
• Run them with one command
• Work as a server for other programs

brew install ollama

Verify:

ollama --version

Should show: ollama version 0.x.x

Start the Ollama server

After installation, start the background server:

ollama serve

You should see:

[GIN] ... Listening on 127.0.0.1:11434

Keep this terminal window open. Open a new terminal window (Cmd+N) for the next steps.

Verify the server is running:

curl http://localhost:11434/api/version

You should see: {"version":"0.x.x"}

Auto-start (convenient)

After installation, the Ollama app appears in your Applications folder. Search for “Ollama” via Spotlight or in Applications and launch it once — it will automatically start the server on every boot. You will see the Ollama icon in the menu bar.

6. Run your first model

In your new terminal (not the one running ollama serve):

ollama run qwen3.5:4b

What happens:

1. If the model is not downloaded — Ollama starts downloading (shows a progress bar)
2. Model size: ~3.4 GB (like one HD movie)
3. Download time: 1–10 minutes depending on your internet
4. After download, you will see the >>> prompt — the model is ready for conversation.

How to chat with the model

Simply type your questions and press Enter. The model will generate a response. After it finishes, the >>> prompt appears again — you can ask the next question.

What to try asking

>>> Write a poem about a programmer
>>> Explain recursion in simple terms
>>> Write a Python function to check if a number is prime
>>> Translate this to English: "Мне нужно запустить AI-модель локально"
>>> Write a study plan for learning Python for a beginner
>>> /bye  # exit the chat

Chat commands

• /bye — exit (model unloads from RAM)
• /clear — clear history (start fresh)
• /show info — model info
• Ctrl+C — stop generation

Try different models

ollama run qwen2.5-coder:7b    # coding model
ollama run llama3.3:8b         # Meta model
ollama run phi4-mini           # small but smart

Model comparison

On a MacBook with 16 GB RAM — all fit. With 8 GB — use only qwen3.5:4b or phi4-mini.

Where to find other models

Browse the full model catalog: ollama.com/search

Each model page shows its size (in the Size column) — use this to check if it fits your RAM.

Download without running

ollama pull qwen3.5:9b    # download only, dont start chat
ollama list               # see all downloaded models

7. Install LM Studio (no terminal alternative)

If you prefer a GUI over the command line, use LM Studio — a desktop app with buttons and menus where everything is done with a mouse.

Installation

1. Open lmstudio.ai in your browser
2. Click “Download for macOS” — downloads a .dmg file
3. Open the downloaded file (usually in Downloads folder)
4. Drag LM Studio to your Applications folder
5. Open LM Studio via Launchpad or Spotlight

First run

1. Search for “Qwen 3.5 4B”
2. Click Download
3. Wait for download (shows progress)
4. Click Load — model loads into memory
5. Click Chat — chat window opens

When to use which

8. Whats next

Tip: Bookmark this page or save it in your browser — you will come back to it when setting up new models.

In section: getting-started · running-models · models · catalog · quantization · memory-and-context · tools · advanced-setup · troubleshooting · apple-silicon
Related sections: Zero Level · AI Agents · Use Cases
Navigation: ← Local Models · ↑ Back to main · 🇷🇺 Русский