Ollama Download: Run Open AI Models on Your Own Computer

If you want to use a chatbot without an internet connection and without sharing what you type with anyone, Ollama is the easiest place to start. It takes care of downloading open models, loading them into memory and running them; all you do is pick a model. With roughly 185,000 stars on GitHub as of October 2026, it is one of the most-followed projects in open-source AI.

What does Ollama do?

• Local models: Download and run open models such as Gemma, Qwen, DeepSeek and gpt-oss from the library at ollama.com/library with a single command.
• Privacy: Because the model runs on your computer, your questions and documents never leave it, and once a model is downloaded you no longer need the internet.
• Local API: In the background it serves a REST API at http://localhost:11434. Interfaces such as Open WebUI, code editors and your own programs connect to this address.
• Libraries: There are official client libraries for Python and JavaScript.
• Developer tool integrations: Ready-made integrations let you run coding assistants on local models.

Who is it for?

It suits people who are curious about AI but don’t want to send data to cloud services, developers building apps on top of local models, and anyone working with limited internet access. If the command line doesn’t scare you, you can have your first model running in a few minutes.

How to install

The easiest way is to grab the installer for your operating system from ollama.com/download, which the Download button opens (OllamaSetup.exe for Windows, Ollama.dmg for macOS). If you prefer the command line, the official install scripts work too.

macOS and Linux:

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

Windows (PowerShell):

irm https://ollama.com/install.ps1 | iex

For Docker users, there is also an official ollama/ollama image on Docker Hub.

First steps

After installing, open a terminal and type the command below. The model downloads the first time, then a chat starts right away:

ollama run gemma4

Use the ollama list command to see which models are installed. If you type just ollama, it offers to run a model or connect to a supported app.

Hardware requirements

• Windows: Windows 10 22H2 or newer. NVIDIA cards need driver 551.61 or later; AMD Radeon cards need a driver with ROCm or Vulkan support.
• macOS: macOS 14 Sonoma or newer. GPU acceleration works on Apple M-series chips; Intel-based Macs use the CPU only.
• NVIDIA support: Cards with compute capability 5.0 and above are supported, a wide range from the GTX 750 Ti to the RTX 50 series.
• Disk: The program needs at least 4 GB. Models take up extra space, ranging from a few GB to hundreds of GB.

It also works without a graphics card, but responses will be noticeably slower. Rather than large models that don’t fit in your computer’s memory, start with smaller versions.

License

Ollama is a free, open-source project under the MIT license. Each model has its own license, so check the license of the model you choose before using it commercially.