Run AI Models on Your Own Computer with Ollama (Step by Step)

By Naresh · October 23, 2026 · Tutorials

Everything AI-related pushes you toward a subscription. But a normal laptop can run a capable language model entirely on your own machine — no API key, no monthly fee, and nothing leaves your computer. The easiest way in is Ollama, a free open-source tool that downloads and runs models with one command.

What you need

Mac, Windows or Linux
RAM is the real limit: as a rule of thumb, a 7-8B parameter model wants roughly 8 GB of free memory, bigger models proportionally more
No GPU required — a GPU (NVIDIA or Apple Silicon) just makes it noticeably faster

Step 1: Install

Download Ollama from ollama.com/download and run the installer. Then confirm in a terminal:

ollama --version

Step 2: Pull a model and chat

ollama pull llama3.1
ollama run llama3.1

The first pull downloads a few GB; after that you are in a chat prompt. Type, get answers, type /bye to exit. See what you have installed with ollama list, remove one with ollama rm llama3.1.

Which model should you pull?

General use: Llama 3.1 (8B for lighter machines)
Coding help: Qwen2.5-Coder or Codellama
Small and fast: Gemma 2 or Mistral variants

Ollama keeps a model library at ollama.com/library — sizes and memory needs are listed there, and new models appear often, so check it for the current options.

Step 3: Use it from other apps

Ollama runs a local server on localhost:11434 with an OpenAI-compatible API. That means the tools you already use can point at it: set the API base URL to http://localhost:11434/v1 and any key, and your app talks to the local model instead of a paid one.

For a ChatGPT-style web interface, Open WebUI is the popular choice — install it with Docker and it connects to Ollama automatically.

Good to know

Private by default: prompts never leave your machine after the model is downloaded.
Works offline: once pulled, no internet needed.
Honest limits: an 8B model on a laptop is handy for summaries, drafts and code explanation, but it is not frontier-level — for hard tasks you still want a big cloud model.

Need something like this built? I take freelance projects in streaming, web apps and automation. Get in touch.