Llama 3.2 on your phone: How to run Meta’s AI model locally

If you have been waiting to try LLama 3.2, Meta’s latest AI model, here’s how you can try it on your smartphone.

Llama 3.2 was introduced to the world during the Meta Connect event that was held in late September. (Image: Meta)

An AI chatbot can come in handy for a variety of tasks. The most popular chatbots today are OpenAI’s ChatGPT and Google’s Gemini, both of which can be used on the web and as smartphone apps. However, last month, Meta came out with its latest AI model, Llama 3.2, which the company describes as state-of-the-art in their class for on-device use cases such as summarisation, instruction following, and rewriting tasks running locally at the edge.

Remember, on-device AI means you could have an AI assistant on the go even without having internet access on your smartphone. If you have been wondering about how to access the chatbot, or perhaps how to get locally on your device. Here’s how to run Llama 3.2 locally on your phone.

How to run Llama 3.2 on your smartphone locally?

Go to the App Store and look for PocketPal AI which has been listed as a productivity app. This app is free and it directs users to the chat with an AI model of their preference. The app listing claims that it offers secure and private conversations with LLMs.

Also Read | Meta AI gets new voice mode, can edit images and dub Instagram reels

Once you have downloaded the app, open it, tap on the left menu, and select models. Now, under Models choose ‘Llama’ and download – “llama-3.2-3b-instruct q4_k” which is 2.02 GB. It may take a while if you are attempting to download it on mobile internet. We recommend trying super-fast WiFi connectivity. Once downloaded, tap on Load to activate the model. Now, return to the main menu, and select Chat to start talking to the AI.

A glimpse of PocketPal AI app from App Store.

PocketPal AI app also offers Google’s Gemma, Microsoft’s Phi, H2O AI’s Danube, and the Qwen model by Hugging Face.

Llama 3.2 is Meta’s in-house open-source model that can help developers create advanced AI models. The model will allow developers to create AI tools that can analyze and explain long documents within minutes.

The above steps have been compiled from a thread on X by Alvaro Cintas, a professor who believes in educating about AI, cybersecurity, and technology.

Tags:
artificial intelligence