Installing Ollama and running your first local model on macOS

Published in the Local LLM Guide • No external cloud required

Running Large Language Models (LLMs) locally offers complete privacy, works without an internet connection, and gives you direct control over open-source models. With Ollama, this has become surprisingly easy on macOS. This guide will walk you through the installation process step-by-step and show you how to communicate directly with your own local model via the terminal.

Also explore our central LLM Hub for an overview of the latest developments and hardware architectures.

Hardware Requirements and System Requirements

Before you begin, it is important to check if your Mac is suitable for running language models locally. Since models run entirely in memory (RAM or Unified Memory), memory capacity is the main bottleneck:

Apple Silicon (M1, M2, M3, M4): Highly recommended. The integrated GPU shares the memory (unified memory), which ensures excellent performance.
Memory (RAM): For smaller models of 7B to 8B parameters (such as Llama 3 or Mistral in 4-bit/8-bit quantization), you need at least 8 GB of RAM; 16 GB or more is highly recommended for smooth operation.
Intel Macs: Supported, but generally perform significantly slower unless a dedicated graphics card is present.

Step 1: Downloading and installing Ollama

Ollama provides a ready-to-use application for macOS that runs as a service in the background and sets up a local API.

Go to the official Ollama website (ollama.com) and download the macOS installer.
Open the downloaded archive and drag the Ollama application to your Applications folder.
Start the application from your Applications folder. You will temporarily see a small llama icon appear in the menu bar at the top of your screen, indicating that the service is active.

Step 2: Verifying via the Terminal

To check if Ollama is functioning correctly and if the command-line tool is accessible, open the default Terminal app (or iTerm2) on your Mac.

Type the following command to check the installed version:

ollama --version

Tip: If the terminal reports that the command is not recognized, make sure you have started the Ollama app at least once and that the CLI symlinks are correctly configured via the app's settings.

Step 3: Downloading and starting your first model

Ollama makes it possible to retrieve an open-source model with a single command and immediately start an interactive chat session. We use Llama 3 as an example here.

Enter the following command in your terminal:

ollama run llama3

The terminal will automatically download the model (this may take a while depending on your internet connection, as the model is several gigabytes in size). Once the download is complete, your prompt will change and you can start chatting immediately:

>>> Write a short greeting in English.
Hello! How can I help you today?

Useful Terminal Commands

While working with Ollama, you can use the following commands:

/bye or Ctrl + D: Closes the active chat session.
/show info: Shows technical details about the active model.
ollama list: Shows an overview of all models stored locally on your Mac.
ollama rm [modelname]: Removes a local model to free up disk space.