← Daily AI recipesStudy✓ Verified Aug 13, 2026

Run Meta's language model on your computer - no internet, subscriptions, or cloud

Install Ollama, download the open-source Llama model, and start chatting in 5 minutes of hands-on work

  1. 1.Download Ollama from the official website and install it like any regular program. On macOS - drag the icon into Applications; on Windows - run the installer and click Next. After installation, the app runs in the background.

    Ollama is a free open-source program. It downloads and runs language models directly on your computer, without sending data to the internet. An 8-billion-parameter model needs at least 8 GB of RAM; for comfortable use - 16 GB.

    Ollama ↗About this tool
  2. 2.Open Terminal: on macOS press Cmd+Space and type 'Terminal'; on Windows press Win+R, type 'cmd', and press Enter. You can also use PowerShell.

    The terminal is a text-based program for controlling your computer. Don't be intimidated by the black window: you'll only need one command. If Ollama is installed correctly, the command ollama --version will show the version number.

  3. 3.Enter the command below and press Enter. On first run, Ollama will download the model (about 4.7 GB), then automatically open the chat. The download takes several minutes depending on your internet speed.

    If you have limited RAM, replace llama3.1:8b with llama3.2:3b (needs ~4 GB) or llama3.2:1b (needs ~2 GB). To see all available Meta model versions, run: ollama search llama

    Ollama ↗About this tool
    Prompt
    ollama run llama3.1:8b
    What this prompt doesThis command downloads Meta's Llama 3.1 model (8 billion parameters) and immediately opens a chat in the terminal. Once the download finishes, a >>> prompt appears - you can start writing messages to the model.
  4. 4.When the >>> prompt appears, the model is ready. Type your request in English or Russian and press Enter - the response will appear right in the terminal. Try this prompt to test it:

    To exit the chat, type /bye or press Ctrl+D. The model stays on disk - next time it will launch instantly. To download without opening a chat, use ollama pull instead of ollama run.

    About this tool
    Prompt
    You are a helpful assistant. Explain recursion in simple terms with a Python example. Keep your answer under 10 lines.
    What this prompt doesThe prompt sets the model's role and a specific task - explaining recursion with a code example. Replace the topic with anything you need: help with text, code, translation, or ideas. The Llama 3.1 model understands Russian well.
  5. 5.For a convenient ChatGPT-style interface, install Chatbox - a desktop app that connects to Ollama. Download the installer, launch the app, in settings select Ollama as the provider and the llama3.1:8b model.

    Chatbox gives you a familiar interface with chat history, text formatting, and saved conversations. The app is free and runs locally. Alternatives: LM Studio (lmstudio.ai) or Open WebUI (requires Docker).

    Chatbox ↗
💰 What it costs to followprices as of 2026-08
  • freeOllamaFully free open-source program, with no limits on the number of requests or models.
  • freeLlama 3.1 model (Meta)The model is free to download. The 8B version needs at least 8 GB of RAM, the 3B - about 4 GB, and the 1B - about 2 GB.
  • freeChatboxFree open-source desktop app with no restrictions. Runs locally; no data is sent to servers.
Try it and be the first to check in

AI services change fast - interfaces and free limits may differ from what's described.