Ollama Setup Guide for Complete Beginners (2026 Edition)

You have heard that you can run AI models on your own computer, completely offline, with no subscriptions and no data sent to the cloud. You are interested, but the instructions you have found online assume you already know what a "quantized 4-bit GGUF" is or how to compile from source.

This guide assumes you know nothing. If you can install an app on your computer, you can follow these steps. By the end, you will have a fully functional AI assistant running on your machine — one that works without the internet and never shares your conversations with anyone.

What Is Ollama?

Ollama is a free, open-source application that downloads, manages, and runs AI language models on your computer. Think of it as a simple layer between you and the AI model — it handles all the technical complexity so you just type a command and start chatting.

It runs on Mac, Windows, and Linux. It supports dozens of models. And it has become the standard tool for running local AI in 2026.

Before You Start: Check Your Hardware

Local AI runs on your computer's resources — primarily RAM. Here is what to expect:

| Your RAM | What You Can Run | Quality Level |

|---|---|---|

| 8 GB | Small models (3B parameters) | Good for basic tasks, writing, simple Q&A |

| 16 GB | Medium models (7B-14B parameters) | Great all-rounder, handles most tasks well |

| 32 GB | Large models (30B-70B parameters) | Near cloud-quality for many tasks |

| 64 GB+ | The biggest open models | Frontier-adjacent quality |

Not sure about your RAM?

Mac: Apple menu → About This Mac → look for "Memory"
Windows: Right-click the Start button → System → look for "Installed RAM"
Linux: Open a terminal and type free -h

If you have 8 GB, do not worry — you can absolutely run meaningful AI models. You will just use smaller ones. If you have 16 GB or more, you are in excellent shape.

A note on Apple Silicon: If you have a MacBook or Mac with an M1, M2, M3, or M4 chip, you are in a particularly good position. Apple Silicon shares memory between the CPU and GPU, which means local AI runs exceptionally well on Macs — often better than on Windows PCs with significantly more expensive hardware.

Best laptop for local AI

macbook-air

Learn More

Step 1: Install Ollama

Mac

Option A — Download the app (easiest):

Go to ollama.com
Click "Download for macOS"
Open the downloaded file and drag Ollama to your Applications folder
Open Ollama from Applications — it will appear as a small icon in your menu bar

Option B — Terminal install:

```

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

```

Windows

Go to ollama.com
Click "Download for Windows"
Run the installer — click through the prompts
Ollama runs as a background service automatically after installation

Linux

Open a terminal and run:

```

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

```

This works on Ubuntu, Debian, Fedora, Arch, and most other distributions. The script detects your system and installs accordingly.

Verify the Installation

Open a terminal (Terminal on Mac, Command Prompt or PowerShell on Windows, any terminal on Linux) and type:

```

ollama --version

```

You should see something like ollama version 0.6.x. If you see a version number, you are ready to go.

Step 2: Download Your First Model

This is where it gets fun. One command downloads a complete AI model to your machine.

For your first model, we recommend Llama 3.2 3B if you have 8 GB of RAM, or Llama 3.1 8B if you have 16 GB or more. These are Meta's open-source models and they are excellent starting points.

For 8 GB RAM:

```

ollama pull llama3.2:3b

```

For 16 GB+ RAM:

```

ollama pull llama3.1:8b

```

The download will take a few minutes. You will see a progress bar. The models are typically 2-5 GB in size. Once downloaded, they live on your hard drive permanently — you only download once.

What Just Happened?

You downloaded a complete AI language model — the same type of technology that powers ChatGPT and Claude — directly onto your computer. It will never need to contact a server. It will never share your data. It is yours.

Step 3: Run Your First Conversation

Start chatting with your model:

```

ollama run llama3.1:8b

```

(Replace with llama3.2:3b if that is what you downloaded.)

You will see a prompt that looks like this:

```

>>> Send a message (/? for help)

```

Type anything. Ask a question. Give it a task. Try these:

```

>>> Explain quantum computing to me like I'm 15 years old

```

>>> Write a professional email declining a meeting invitation

```

>>> What are the pros and cons of starting a small business in 2026?

```

The model will respond in seconds. You are now having a conversation with an AI that is running entirely on your hardware. Disconnect from Wi-Fi if you want to prove it — it keeps working.

To exit the chat, type /bye.

Step 4: Explore More Models

Ollama gives you access to dozens of models, each with different strengths. Here are the ones worth trying:

Best General-Purpose Models

|---|---|---|---|

Specialist Models

| Model | Best For | Command |

|---|---|---|

| DeepSeek Coder V2 | Programming and code generation | ollama pull deepseek-coder-v2 |

| Llama 3.2 Vision | Understanding images (describe, analyze) | ollama pull llama3.2-vision |

| Phi-3 Mini | Very fast responses on low-end hardware | ollama pull phi3:mini |

You can have multiple models installed and switch between them. Each ollama run [model] starts a new session with that model.

To see all models you have downloaded:

```

ollama list

```

To remove a model you no longer want:

```

ollama rm [model-name]

```

Step 5: Set Up Open WebUI (The ChatGPT-Like Interface)

Chatting in a terminal is functional but minimal. Open WebUI gives you a beautiful, browser-based interface that looks and feels like ChatGPT — complete with conversation history, model switching, dark mode, and file uploads.

Install With Docker

Open WebUI runs via Docker, which is a tool for running applications in containers. If you do not have Docker:

Download Docker Desktop from docker.com (free for personal use)
Install it and make sure it is running (you will see a whale icon in your menu bar or system tray)

Then run this single command:

```

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:main

```

Wait for the download to complete (first time takes a few minutes), then open your browser and go to:

```

http://localhost:3000

```

You will be asked to create an account. This account is stored only on your machine — it is not sent anywhere. Create a username and password, and you are in.

What Open WebUI Gives You

A clean, dark-mode chat interface
Conversation history that persists across sessions
The ability to switch between any Ollama model mid-conversation
File upload and document analysis (paste a PDF and ask questions about it)
Custom system prompts — tell the AI to behave as a writing assistant, code reviewer, or anything else
Multiple user accounts if you share a computer

It automatically detects all the models you have installed in Ollama. No configuration needed.

Troubleshooting Common Issues

"Command not found" after installing Ollama:

Close your terminal and open a new one. On Mac, you may need to restart your computer. On Windows, make sure you ran the installer as Administrator.

Model download is slow:

The models are several gigabytes. On a slower connection, the initial download can take 15-30 minutes. Once downloaded, it is saved locally — you never need to download it again.

Responses are very slow:

You may be running a model too large for your hardware. Try a smaller model. If you have 8 GB of RAM and are running a 14B model, switch to the 3B version.

Open WebUI shows "no models available":

Make sure Ollama is running in the background. On Mac, check for the Ollama icon in the menu bar. On Windows, check the system tray. If it is not running, open the Ollama app.

Docker command fails:

Make sure Docker Desktop is installed and running. You should see the Docker whale icon in your menu bar or system tray. If you just installed it, restart your computer.

What To Do Next

You now have a local AI assistant. Here are three things to try this week:

Replace one ChatGPT session per day with your local model. Notice what it handles well and where it falls short.
Try a specialist model — if you write code, grab DeepSeek Coder. If you work with documents, use Open WebUI's file upload.
Disconnect from the internet and use the AI offline. This is the moment it clicks — you have real AI capability with absolute privacy.

The models are improving fast. What required a high-end gaming PC in 2024 now runs on a MacBook Air. And every conversation you have stays on your machine, forever.

Frequently Asked Questions

Is Ollama free to use?

Yes, Ollama is completely free and open-source. There are no subscriptions, usage caps, or payment tiers — you download it once and run it as many times as you like. The models you run through Ollama are also free, open-weight models published by Meta, Mistral, and others.

Does Ollama work without an internet connection?

After the initial download of Ollama and your chosen model, the entire system works offline. Your prompts never touch a network connection — they go from your keyboard to your CPU or GPU and back. You can disconnect from Wi-Fi entirely and every conversation will work exactly the same.

Is Ollama safe to download and install?

Ollama is maintained by a funded, venture-backed company with public founders and an open-source codebase available for anyone to audit on GitHub. It does not call home with your conversations. The only network activity is downloading models and updating the software when you choose to update. It is widely used by developers and security researchers who have audited it extensively.

What is the difference between Ollama and ChatGPT?

ChatGPT runs on OpenAI's servers and every prompt you type is sent over the internet to their systems. Ollama runs an AI model directly on your own computer — nothing leaves your device. ChatGPT gives you access to GPT-4 and similar frontier models; Ollama runs open-weight models like Llama and Mistral that are somewhat behind the frontier but capable for most everyday tasks.

How much storage space do Ollama and its models require?

Ollama itself is under 200 MB. Models range from about 2 GB for small 3B-parameter models to 40+ GB for the largest 70B models. A good starting setup — Ollama plus a 7B to 8B model — requires roughly 5–7 GB of storage. You can delete models you no longer use with ollama rm [model-name] to reclaim space at any time.

Does Ollama work well on Apple Silicon Macs?

Yes — Apple Silicon is one of the best platforms for local AI. The unified memory architecture on M1, M2, M3, and M4 chips means the GPU and CPU share the same fast RAM pool, which dramatically improves local model performance. A MacBook Air M2 with 16 GB of RAM outperforms many dedicated gaming PCs for local AI workloads at a fraction of the power draw.

Get The Private Intelligence — weekly

AI tools, privacy guides, and workflow tips for people who want to use AI powerfully without giving up their data. Join 1,000+ readers.

Affiliate Disclosure: This article may contain affiliate links. If you make a purchase through these links, we may earn a small commission at no extra cost to you. We only recommend products we genuinely believe in. This helps support our work and allows us to continue providing free content.