Gemma 4 E4B IT (LiteRT-LM)

Gemma 4 E4B IT (LiteRT-LM)

Runs a small Gemma 4 model on phones, desktops, IoT boards and the web without an internet connection. A 3.66 GB LiteRT-LM build for developers who want private on-device generation.

Made by litert-community on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Gemma 4 E4B Instruct packaged in the LiteRT-LM format, so it can run on Android, iOS, desktop, IoT devices and the web. It is a small Gemma 4 model meant for on-device use. Running it locally means people can use generative AI privately, with no internet connection.

Who it suits: Mobile and embedded developers who want a compact model inside their own app, and teams that need text, image or audio features to work offline. Google's AI Edge Gallery app runs on the same stack.

Industry | Who uses it | How they use it | Example
Mobile apps | Android and iOS developers | Ship an offline assistant inside an app | A chat helper that works without signal
Embedded and IoT | Device makers | Run a language model on small hardware | A smart device that answers spoken questions
Privacy-focused products | Product teams | Keep user text on the device | A notes app that summarizes locally
Research | Edge AI researchers | Measure memory and speed on small devices | Comparing memory use across platforms

You can skip it if you want the largest model quality available, since this one is built to stay small enough for devices.

Where to use it:

• Android and iOS apps, including Google's AI Edge Gallery
• Desktop computers and IoT devices
• The web, through LiteRT-LM

Use case | Who | What they get back
Offline assistant | App developers | Text replies generated on the device
Image and audio features | Product teams | Answers that use a picture or a sound clip
Low-memory deployment | Device makers | A model that loads parts only when needed

How to set it up:

• Python, to install the Hugging Face client
• A LiteRT-LM runtime in your app or on your device
• About 3.66 GB of space for the model file

1. Install the Hugging Face client.
2. Download the model files to your computer.
3. Load the model file with LiteRT-LM in your app.

pip install -U huggingface_hub
hf download litert-community/gemma-4-E4B-it-litert-lm

How to use it: Load the model file through LiteRT-LM, then send prompts and read the replies, all on the device. The source shows no code, so follow the LiteRT-LM guide for the calls. The vision and audio parts load only when a request needs them.

Tool | What it does
LiteRT-LM | Runs the model file and adds KV-cache handling, prompt templates and function calling
LiteRT | Supplies the hardware acceleration, XNNPack on CPU and ML Drift on GPU
AI Edge Gallery | Showcase app built on the same stack

Example: You build a travel notes app and ask the model on the phone to tidy a few lines you dictated. It does the work with no network, and the text never leaves the handset.

For newcomers and experts: If you are new, try Google's AI Edge Gallery app first to see a model like this run on a phone. Then follow the LiteRT-LM guide to bring it into your own app.

If you know the field: of the 3.66 GB, 2.24 GB is the text decoder and 0.67 GB is the embedding table. The runtime keeps the main weights in memory and memory-maps the embeddings, which saves working memory on some platforms.

Your prompt
LiteRT-LM
XNNPack on CPU or ML Drift on GPU
Reply on device

Good to know: The whole file is 3.66 GB. Vision and audio models load only when needed, which lowers memory use further.

Title: Model file contents in GB
Whole file: 3.66 GB
Text decoder weights: 2.24 GB
Embedding parameters: 0.67 GB

• Platforms: Android, iOS, desktop, IoT and web
• License: Apache 2.0
• Cost to run: none named
• Data it reaches: model files downloaded to your computer, nothing else named
• Not for: the source names none, though it is a small model made for on-device use

On Hugging Face, 13 November 2026

• Library: litert-lm
• Downloads in the last 30 days: 467,257
• Likes: 250
• License: apache-2.0
• Based on: google/gemma-4-E4B-it
• Fine-tunes: 4
• Tags: litert-lm

Links

• Model: https://huggingface.co/litert-community/gemma-4-E4B-it-litert-lm
• Files: https://huggingface.co/litert-community/gemma-4-E4B-it-litert-lm/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download litert-community/gemma-4-E4B-it-litert-lm

Then follow the usage in its model card. Nothing in this kit runs by itself.