Privacy & offline

Is Ollama fully offline? — privacy & local mode guide

Yes. After downloading the installer and pulling a model, Ollama runs with no internet connection. Prompts and responses never leave your machine. This guide explains how offline mode works, where data is stored, and how to enforce offline-only use on Windows.

Is Ollama fully offline? Yes.

Ollama runs entirely on your local machine. After you download the installer and pull a model, Ollama generates text with no internet connection required. Your prompts, responses and conversation history never leave your computer.

The only time Ollama uses the internet is when you explicitly download a new model with ollama pull. Everything else — inference, the local API, the CLI — works completely offline.

You can disconnect from the internet after pulling a model and Ollama will continue to work normally.

Local by design

When you run ollama run llama3, the model weights are loaded from your local disk into RAM or VRAM. The inference runs on your CPU or GPU. Results are returned to your terminal or to localhost:11434 — there is no network call to any external server.

This is fundamentally different from cloud AI services (ChatGPT, Claude, Gemini) where your prompts are sent to remote servers. With Ollama, the entire language model and all computation stays on your hardware.

ActionUses internet?
Running a model (ollama run)No — fully local
Calling the API at localhost:11434No — local only
Downloading a model (ollama pull)Yes — downloads from registry
Checking for updatesOnly if you manually check

Where does Ollama store data on Windows?

All Ollama data is stored locally in your Windows user profile:

PowerShell
# Model weights and manifests:
C:\Users\YourName\.ollama\models
# List all installed models:
PS> ollama list
NAME SIZE
llama3:latest 4.7 GB
mistral:latest 4.1 GB

There is no cloud sync, no telemetry upload, and no data sent to Ollama servers during normal use. Model files stay on your local drive until you explicitly delete them with ollama rm modelname.

How to run Ollama completely offline

Once you have downloaded your models, you can go fully offline:

  • 1

    Pull the models you need while connected

    cmd.exe
    C:\> ollama pull llama3
    C:\> ollama pull mistral
    Both models downloaded and ready
  • 2

    Disconnect from the internet

    Disable Wi-Fi or unplug the ethernet cable. Ollama continues running normally with no errors.

  • 3

    Use Ollama as normal

    cmd.exe — offline
    C:\> ollama run llama3
    >>> Send a message (/? for help)
    # Works with no internet connection

Block Ollama from the internet with Windows Firewall

If you want to enforce offline use and prevent Ollama from ever making outbound connections (for example, in a high-security environment), add a Windows Firewall outbound rule:

PowerShell — admin
PS> New-NetFirewallRule -DisplayName "Block Ollama Outbound" -Direction Outbound -Program "$env:LOCALAPPDATA\Programs\Ollama\ollama.exe" -Action Block
This blocks ALL outbound connections for Ollama, including model downloads. You will need to remove the rule to pull new models.

Privacy and offline questions

Does Ollama send my prompts to the cloud?
No. When you run a model locally, your prompts are processed entirely on your own hardware. Nothing is sent to Ollama servers or any third-party service. The model runs locally and all output stays on your machine.
Does Ollama collect telemetry?
Ollama does not collect telemetry or usage data by default. Check the official Ollama privacy policy for the most current information. This site's privacy policy covers only this website's analytics — see Privacy Policy.
Can I use Ollama in an air-gapped environment?
Yes. Download the installer and model files on a connected machine, transfer them to the air-gapped system, and install Ollama manually. Models are stored as regular files in the .ollama/models folder and can be copied between machines.
Is my conversation history stored anywhere?
Ollama does not persist conversation history between sessions by default. Each ollama run session starts fresh. Within a session, context is held in RAM only and is cleared when you exit. No conversation logs are written to disk unless a third-party client app you are using does so explicitly.
Does ollama require internet after model download?
No. Once a model is downloaded with ollama pull, you can completely disconnect from the internet and Ollama will continue working normally. The only internet connection needed is the initial model download.

Want to try Ollama offline?

Download and install Ollama, pull a model, then disconnect.

Download Ollama