Start from zero

Ollama, explained for a complete beginner.

Ollama is an app that lets you download AI models and run them on your own computer—almost like installing different AI brains.

01

The one idea to remember

Ollama is the engine. The model is the AI.

Your computerOllamaAI modelAnswer

Gemma, Qwen, DeepSeek, GPT-OSS, and Llama are model families. Ollama downloads, stores, starts, stops, and connects them to other software.

Model
The AI brain.
Pull
Download a brain.
Run
Start a brain.
Prompt
What you say to it.
API
A way for software to talk to it.
Local
It runs on your computer.
02

Your first ten minutes

Install, choose, run, chat.

  1. Install Ollama.Use the official download for macOS, Windows, or Linux.Open official download ↗
  2. Open Ollama.On macOS and Windows, the desktop app provides a graphical chat experience. Terminal is optional for a beginner.
  3. Choose a sensible model.Check the exact download size on its official page before choosing. Larger is not automatically better for your computer.
  4. Start it.In Terminal, enter the command below. If the model is not installed, Ollama downloads it first.
    ollama run gemma3
  5. Talk naturally.Try: “Explain artificial intelligence like I am ten years old.” Continue the conversation, then exit when finished.
Before downloading:

Model files can range from hundreds of megabytes to hundreds of gigabytes. Confirm the exact tag, file size, memory requirement, and license on Ollama first.

03

Copy & run from this website

One click copies the right command.

Select a model and press Copy & run. The site copies its visible ollama run … command and shows the next steps for macOS, Windows, or Linux.

  1. Install Ollama.Use the official Ollama download for the computer that will run the model.
  2. Check the model.Review its exact tag, download size, license, and hardware needs on the official page.
  3. Open Terminal.Use Terminal on macOS or Linux, or Windows Terminal / PowerShell on Windows.
  4. Paste and press Enter.If the model is not installed, ollama run downloads it first and then starts it.
You approve the final action:

A normal website cannot open, paste into, or execute commands in Terminal. ailocal.click copies only the command you can see. You choose when to paste it and press Enter.

04

What the labels mean

Models are good at different things.

Vision
Understands compatible images and visual input.
Tools
Can request functions, searches, or actions supplied by another program.
Thinking
Supports reasoning controls on compatible requests.
Embedding
Turns text into vectors for search, recommendations, and RAG.
Audio
Accepts audio on supported variants.
Cloud
Runs on Ollama’s servers, not entirely on your computer.

Some vision models can inspect photographs, screenshots, charts, diagrams, and documents. The desktop app can also accept supported files, but document length and context settings affect memory use.

05

The ten useful commands

You do not need to memorize them today.

ollama

Open Ollama’s interactive terminal menu.

ollama run gemma3

Download Gemma 3 if necessary, then start chatting.

ollama pull gemma3

Download a model without starting it.

ollama ls

List models stored on your computer.

ollama ps

See models currently loaded in memory.

ollama stop gemma3

Stop a loaded model.

ollama rm gemma3

Remove a downloaded model and recover its disk space.

ollama signin

Sign in when you want to use Ollama cloud features.

ollama launch

Connect a model to supported applications and coding agents.

ollama serve

Start the local Ollama server manually when needed.

06

A crucial distinction

Local and cloud models are not the same.

Local

Your hardware does the work.

  • Can work offline after downloading.
  • Prompts can stay on your machine.
  • No per-message API charge for local inference.
  • Your memory, storage, and compute set the limit.

Cloud

Ollama’s servers do the work.

  • Requires an internet connection and Ollama account.
  • Runs models too large for many personal computers.
  • Uses cloud infrastructure rather than only local hardware.
  • Policies and availability can change; check Ollama.
07

When software talks to Ollama

localhost:11434 means “Ollama on this computer.”

localhost is your own machine. 11434 is the default local network port Ollama listens on. Together, http://localhost:11434/api is the local API address programs can use.

Your appOllama APILocal model

Ollama also offers official Python and JavaScript libraries and partial OpenAI-compatible endpoints. This is how document tools, editors, agents, and your own applications can use a local model underneath.

08

A calm learning path

Go one level at a time.

  1. Level 1Use AI

    Install Ollama, choose one model, and chat.

  2. Level 2Learn the basic commands

    Run, pull, list, stop, and remove models.

  3. Level 3Try files and images

    Use compatible vision models and understand context limits.

  4. Level 4Connect software

    Learn localhost, APIs, Python, JavaScript, and supported integrations.

  5. Level 5Build systems

    Explore tools, RAG, embeddings, custom models, and coding agents.

  6. Level 6Operate responsibly

    Measure quality, privacy, hardware use, provenance, and production reliability.

Ready to choose?

Start small. Check the size. Keep the official page open.

Primary references: Quickstart, CLI reference, API introduction, and Cloud documentation.

Copy & run

Choose a model

One command downloads this model if needed, then starts it with Ollama.

ollama run
Command ready.

Copy it, then follow the three steps below.

  1. 1

    Open Terminalon the computer where Ollama is installed.

  2. 2

    Paste the commandusing your normal paste shortcut.

  3. 3

    Press EnterOllama downloads the model if needed, then starts it.

The website copies a visible command; your browser cannot open or control Terminal. You choose when to paste and press Enter.

Model files can be large. Check the exact tag, file size, license, and hardware needs first.

Copy & run help for beginners