RUN YOUR FIRST LOCAL AI MODEL

START WITH A SIMPLE CONVERSATION

Install Ollama on Windows, download the Gemma 3 1B text model, and send your first prompt. This walkthrough takes you from a fresh installation to a working local AI conversation, then shows you how to close the session and run the model again.

The goal is to confirm that your basic setup works before adding larger models, extra interfaces, or creative integrations.

This guide uses a small, tested model to teach the setup process. You can try other models once your setup works.

BEFORE YOU BEGIN

CHECK YOUR WINDOWS PC

This walkthrough uses Windows 10 version 22H2 or newer, including Windows 11. You’ll need an internet connection to download Ollama and the model, plus enough free disk space for both.

Allow at least 4 GB for the Ollama installation and additional space for the model. The Gemma 3 1B download is approximately 815 MB, but download size is different from the memory needed to run it.

Close demanding applications before your first test. Performance will depend on your CPU, available system RAM, and any supported GPU. We’ll start with a short conversation to keep the initial workload small.

INSTALL OLLAMA

DOWNLOAD THE WINDOWS INSTALLER

Visit ollama.com/download/windows and download the official Windows installer. Open OllamaSetup.exe and follow the installation prompts.

The standard installer installs Ollama for your Windows account without requiring administrator permissions. Once installed, Ollama runs in the background and makes its commands available in PowerShell.

For this walkthrough, we’ll use PowerShell to download and run a specific local model.

CHECK THE INSTALLATION

OPEN POWERSHELL AND LIST YOUR MODELS

Open the Windows Start menu, search for PowerShell, and launch it. If PowerShell was open during installation, close it and open a fresh window so it can recognize the new command.

Type the following command and press Enter:

ollama ls

You should see a table with columns such as NAME, ID, SIZE, and MODIFIED. An empty table is expected if you haven’t downloaded any models yet.

If Windows says the command is not recognized, reopen PowerShell and try again. If you see a connection error, open Ollama from the Start menu, then repeat the command.

DOWNLOAD AND RUN THE MODEL

START GEMMA 3 1B

In PowerShell, type the following command and press Enter:

ollama run gemma3:1b

On the first run, Ollama downloads the model before loading it. Keep your internet connection active and wait for the download to finish. Download time depends on your connection, and loading time depends on your hardware.

When the model is ready, you’ll see an input prompt beginning with >>>. This is where you type your messages to the model.

The gemma3:1b name selects the small, text-only model used in this walkthrough. Use the full name so you download the intended version.

SEND YOUR FIRST PROMPT

TEST A SHORT CONVERSATION

At the >>> prompt, type the following message and press Enter:

Give me three ideas for organizing a creative project folder. Keep each idea to one sentence.

Wait for the response to finish. Then send a follow-up message:

Turn those ideas into a short checklist.

If the model responds to both messages, you’ve completed your first local AI conversation. The exact wording will vary, and a small model may give imperfect answers. For this test, focus on whether it accepts your prompts and produces readable responses.

Record anything unexpected, such as an error or a long delay, so you can investigate it before expanding your setup.

CLOSE AND RESTART THE SESSION

RUN YOUR MODEL AGAIN WHEN YOU NEED IT

To end the conversation, type the following at the >>> prompt and press Enter:

/bye

You’ll return to PowerShell. To release the model from memory immediately, run:

ollama stop gemma3:1b

This keeps the downloaded model on your disk. When you want to start a fresh conversation, run:

ollama run gemma3:1b

You can reuse the downloaded model without repeating the initial download. The new conversation won’t automatically restore your previous chat, so copy any useful responses into your project notes before exiting.

IF SOMETHING GOES WRONG

CHECK THE SIMPLE THINGS FIRST

If PowerShell doesn’t recognize the ollama command, close it and open a fresh window. If the problem continues, confirm that Ollama finished installing.

If Ollama reports a connection error, launch it from the Windows Start menu and try the command again.

If the model download fails, check your internet connection and free disk space, then retry ollama run gemma3:1b. Check the spelling, including the colon and the 1b ending.

If you receive a memory error, close demanding applications and retry. If responses are slow, keep your prompts short while checking your available RAM and GPU compatibility. A slow response alone doesn’t mean the installation failed.

If the problem persists, copy the exact error message and consult Ollama’s troubleshooting documentation before changing advanced settings.