Skip to content

MobileLLM-125M

A small language model built for on-device use

Visit official site

What MobileLLM-125M is

The model is described as a Meta autoregressive language model using an optimized transformer design for devices with limited resources.

What you can do with it

  • Produce text without sending work off the device
  • Support commonsense reasoning tasks locally

Getting started

  • Find the model on the HuggingFace platform
  • Load pretrained weights with the provided code

Pros & cons

✓ What we like

  • Loads pretrained weights through the HuggingFace platform
  • Supports fine-tuning or direct inference

! What to watch out for

  • Custom training requires fetching the codebase from GitHub
  • Users need to load the pretrained model with provided code

FAQ

Who developed it?

The material names Meta as the developer.

What sizes are offered?

The material lists models from 125M to 1.5B parameters.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review