Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

Llama 3.2: it sees, and it is small enough for a phone

·

Meta releases Llama 3.2: models that understand images, and small text models for phones and laptops.

A working adult reviews a summary on a phone.
Small Llama 3.2 models can run on the user’s own device. Illustration: Denkbaar.

On 25 September 2024, Meta releases Llama 3.2.

What happened

  • There are two models that understand images: 11B and 90B.
  • There are two small text models: 1B and 3B. They fit on phones and other small devices.
  • The small models have a context window of 128,000 tokens. Meta names these tasks: summarising, following instructions and rewriting text, on the device itself.
  • According to Meta, they work from day one on Qualcomm and MediaTek chips, and are optimised for Arm processors.
A tablet shows a recognised box and a phone shows text blocks.
The family includes models for images and small models for text. Illustration: Denkbaar.

Why it mattered

A language model does not always have to run in a data centre. For simple tasks, it can run on the user’s own device.

Hands sort form cards with help from a phone.
For a small fixed task, a local model can be enough. Illustration: Denkbaar.

What you can do with it now

Look for a task in your process that is small and fixed, such as summarising a text or sorting a form. A small local model can be enough there.

Written afterwards on 12 November 2026, from the source below.

Sources

  1. Meta: Llama 3.2: Revolutionizing edge AI and vision with open, customizable models

Topics: open-sourcemodelsmeta