Run a model on your phone

The same small promise as running one on a laptop, but smaller and stranger: the phone in your pocket can run its own AI. No account, no internet once it’s downloaded, nothing sent anywhere. In a few minutes, and a few taps.

If running a model on a computer felt surprising, running one on a phone feels almost silly. Let’s do it.

What you need

• A reasonably recent phone (iPhone or Android).
• About 2GB of free space for a small model.
• Wifi for the one-time download. How long it takes is down to your wifi speed, not the app, so there’s no honest “few minutes” promise here.

Step 1 - Install a local AI app

We’ll use PocketPal AI. It is free, open-source, and does the whole thing on your phone with no account.

Open your app store and search for PocketPal AI:

• iPhone: the App Store.
• Android: Google Play.

Install it like any other app.

Step 2 - Download a small model

Open the app and go to its list of models (usually a “Models” or “Hub” screen). You’ll see a menu of models you can download.

Pick a small one, around 1 to 2 billion (you’ll see sizes like 1B or 2B). Good first choices are names like Gemma 3 1B, Qwen 3 1.7B, or Phi-4 mini. Tap to download.

This is the only step that needs wifi. It downloads once (a gigabyte or two) and then lives on your phone.

Step 3 - Load it and chat

When the download finishes, tap the model to load it, then open a new chat. You’ll get a familiar chat box. Type something:

You:  in one sentence, why would I run AI on my phone?

AI:   So it works anywhere, keeps your conversations private on your
      own device, and never needs an account or a signal.

That reply was written by a model running on your phone. Not in a data centre. In your hand.

The check that proves it

Here is the fun part. Turn on airplane mode - no wifi, no signal, nothing. Now ask it another question.

It still answers.

That is the whole point, made undeniable. There is no server on the other end. The AI is on the phone itself. With everything switched off, there is simply nowhere for your words to go, and it answers all the same.

A few honest notes

• It will be slower than a laptop or a cloud app, and the small models are less clever. That is normal. You are running AI on a phone, which is a small miracle in itself.
• Bigger models will struggle or refuse. Stick to the 1B-2B range on a phone. If it gets hot or sluggish, pick a smaller one.
• Your phone may already have one built in (recent phones ship with a small on-device model), but the app above is the reliable way to try it yourself today, on almost any phone.

If something looks wrong

• The download is slow or seems stuck - that is your wifi, not the app. It is a big file. Stay on wifi, keep the app open, and let it finish.
• The app crashes, the model won’t load, or the phone gets hot - the model is too big for your phone. Delete it and pick a smaller one (aim for 1B).
• You can’t find the model I named - model lists change often. Any small one labelled around 1B or 2B will do; the size matters more than the exact name.
• Replies come slowly - normal on a phone, especially an older one. A smaller model replies faster.

What “working” looks like: the download reaches 100%, the model loads, and a reply appears in the chat. The real proof is the airplane-mode test below.

Wow, on a phone

You just ran an AI, offline, on the device in your pocket. A year ago most people would have told you that was impossible.

If you have not yet done it on a computer, that path is even simpler and the models are smarter - see run your first model. And if you want to understand what you just ran, what a local LLM is explains it in plain words.