Open Weight Models · Lesson 2

Install Ollama and check your setup

Prepare your computer and verify that your local Ollama service is ready.

What you will learn

Install Ollama using the official instructions for your operating system.
Distinguish the command-line client from the background service.
Verify that the client is installed and can reach the local service without downloading a model.
Recognise an empty model list as a valid starting state.

Prepare your computer

Choose the instructions for the operating system you are actually using. On a managed classroom or work computer, follow the software installation process for that device. Use Ollama's official download site and read the linked platform requirements before starting; hardware and operating-system support can change.

There are two parts to recognise. The terminal command sends requests, while a background service performs the work. A working command does not necessarily mean that its service is running. In this lesson, install the runner and check the connection; model downloads come next.

Keep the hardware and free-space notes from Lesson 1.
Close unnecessary applications before the later inference exercises.
Use a normal terminal: PowerShell on Windows or Terminal on macOS/Linux.
The local course workflow does not require signing into a cloud model or creating an API key.

Windows: install the application, then reopen PowerShell

Open the official download page, select Windows, and run OllamaSetup.exe. Follow the installer and launch Ollama. The normal installer makes the ollama command available to your account and runs the application in the background.

Close any terminal that was open before installation, then open a new PowerShell window. Run the verification commands below. If PowerShell cannot find ollama, first check that installation completed and that you reopened the terminal; this is different from an error connecting to the service.

Use the platform guide

The Windows documentation lists current operating-system, storage, and GPU requirements. You do not need to install Docker or WSL for this course's native Windows route.

macOS: install the app and enable its command

Download the macOS package from the official site. Open the disk image and move Ollama into Applications. Launch the application and follow its startup instructions, including the prompt to make the command-line tool available if it is not already on your PATH.

Open a new Terminal window for verification. Check the current macOS requirements in the platform guide: support for an operating system and support for GPU acceleration are separate questions. If the app is running already, do not start an additional server just to run the checks.

Linux: understand the installer before running it

Follow the official Linux guide for your distribution, architecture, and hardware. Its standard installer is a shell script. The equivalent staged approach below downloads the script so you can read it before execution. Installation changes the system and may request administrator privileges.

Run the first command, inspect the file with less, and press q to leave the viewer. Execute the final command only when you are ready to install. If your environment requires a manual installation, use the architecture-specific instructions in the official guide instead.

Linux terminal: download, inspect, then install · bash

curl -fsSL https://ollama.com/install.sh -o ollama-install.sh
less ollama-install.sh
sh ollama-install.sh

Starting the service

If installation has not started a service, follow the Linux service instructions. For a manual foreground session, run ollama serve in one terminal and keep it open while you use another. Do not launch a second server when one is already listening.

Verify the client and the service separately

Run these commands in your normal terminal, one at a time. The version command reports the installed client; the list command also checks that the client can communicate with the configured service. A model table with no entries is expected on a fresh installation.

If the command is missing, revisit installation and PATH setup. If the service cannot be reached, open the desktop app or start the Linux service. If a manual server reports that its address is in use, check whether Ollama is already running rather than starting repeated copies.

Verify the client and the service separately
Observation	Meaning	Next step
A version is printed	The shell can find the client	Check the service with ollama list
A model table appears	The service answered	Proceed even if there are no model rows
A connection error appears	The service or connection needs attention	Check the app/service and retry

PowerShell, macOS Terminal, or Linux shell · text

ollama --version
ollama list

Put it into practice

Record a clean setup check

Install using only the section for your operating system and the corresponding official guide.
Open a fresh terminal and run ollama --version. Record the reported version and any warning separately.
Run ollama list. Record whether the service returned a table and whether it contains any existing models.
If a check fails, record the exact message and resolve the relevant installation or service issue before continuing.

You have completed this task when…

The client version is recorded, and ollama list returns without a connection error.
You can identify how the service is started on your platform.
You understand that an empty list is a successful setup check, not a missing-model error to fix immediately.

Official documentation

Use these references for platform requirements, current options, and further detail.