How to Install Ollama on Windows, macOS, and Linux

Ollama lets you download and run large language models like Llama 3, Mistral, and Gemma directly on your own computer. You get full privacy because your prompts never leave your machine, you pay nothing per token, and the models keep working offline once you download them. This guide walks you through the exact installation steps for each operating system, then shows you how to verify the setup and run your first model.

System Requirements Before You Install

Check your hardware before you start so you avoid slow performance later.

RAM: 8 GB minimum, 16 GB recommended for a smooth experience.
Disk space: The Ollama app uses around 300 MB. Each model needs more. A 7B model takes about 4 GB, a 34B model around 19 GB, and a 70B model close to 39 GB.
Storage type: Use an SSD. Models load far faster from solid state storage.
GPU (optional): Ollama runs on CPU alone, but a GPU speeds things up. NVIDIA cards use CUDA, AMD cards use ROCm, and Apple Silicon uses Metal. Ollama bundles the CUDA components it needs, so you do not install the CUDA toolkit separately.
Internet: You only need a connection for the initial app and model download.

Apple Silicon Macs handle models especially well because the CPU and GPU share the same memory pool. A 32 GB M-series Mac can load models that would otherwise need a dedicated graphics card.

How to Install Ollama on Windows

Windows 10 and Windows 11 both support Ollama with a standard installer.

Open your browser and go to the official download page at ollama.com/download.
Click the Windows button to download the .exe installer file.
Double click the downloaded file to launch the installer.
Click through the prompts the same way you would for any other Windows app. The installer adds Ollama to your PATH automatically.
Wait for the installer to finish. Ollama then runs in the background and places an icon in your system tray.

If the installer fails to launch, right click the file and choose Run as administrator. Some antivirus tools flag new installers, so pause your antivirus briefly during the install if it blocks the process.

How to Install Ollama on macOS

You have two clean options on macOS 11 or later.

Option 1: Direct download

Visit ollama.com/download in your browser.
Click the macOS button to download the .zip file.
Open the downloaded file and drag the Ollama app into your Applications folder.
Launch Ollama from Applications. Approve any permission prompt macOS shows on first run.

Option 2: Homebrew

If you already use Homebrew, run this single command in Terminal:

brew install ollama

Homebrew downloads and installs Ollama and adds the ollama command to your shell. Apple Silicon Macs (M1 through M4) accelerate models through Metal right away. Intel Macs run Ollama too, though they process tokens more slowly.

How to Install Ollama on Linux

Linux users install Ollama with the official one line script. Open a terminal and run:

curl -fsSL https://ollama.com/install.sh | sh

The script downloads the Ollama binary, adds the ollama command to your PATH, and creates a systemd service called ollama. That service starts automatically and relaunches after a reboot. The installer needs root access to set up the systemd service, so it prompts you for your password.

Ubuntu 20.04 and later work well. If you prefer a containerised setup, pull the official Docker image instead of running the script.

Verify Your Installation

Confirm Ollama installed correctly before you pull a model. Open a terminal or PowerShell window and run:

ollama --version

The command prints the installed version number. If you see the version, your installation works.

Run Your First Model

Pull and run a model with one command. This example downloads and starts Llama 3:

ollama run llama3

The first run downloads the model, which takes a few minutes depending on your connection. After the download finishes, Ollama drops you into an interactive chat prompt. Type a question and press Enter to get a response. Type /bye to exit the chat.

You can also pass a prompt directly:

ollama run llama3 "Explain quantum computing in simple terms"

Useful Ollama Commands

These commands cover daily use once you have Ollama running.

ollama list                # Show every model you installed
ollama pull mistral        # Download a model without running it
ollama run llama3          # Start an interactive chat
ollama rm llama3           # Remove a model to free disk space
ollama stop                # Stop the running model

Remove models you no longer use with ollama rm. Large models take significant disk space, and clearing them keeps your drive healthy.

Fix Common Installation Problems

The installer will not run on Windows: Right click the .exe file and choose Run as administrator. Pause your antivirus during the install if it interrupts the process.
The ollama command is not found: Close and reopen your terminal so it picks up the updated PATH. On Linux, confirm the install script finished without errors and rerun it if needed.
Models load slowly: Move Ollama and your models onto an SSD, add more RAM, or use a supported GPU. CPU only setups run smaller models at roughly 7 to 12 tokens per second, which suits experimentation more than daily work.
The service does not start on Linux: Check the systemd service status with systemctl status ollama and restart it with sudo systemctl restart ollama.

Final Setup Tips

Start with a smaller model like a 7B or 8B parameter version so you confirm everything works before you move to heavier models. Keep an eye on disk space as you add models, and remove the ones you stop using. Once your installation runs cleanly, you have a private, offline AI environment ready for text generation, coding help, and question answering, all without subscriptions or per token fees.

Read more: