Gemma 3n fully available in the open-source ecosystem!

Gemma 3n was announced as a preview during Google I/O. The on-device community got really excited, because this is a model designed from the ground up to run locally on your hardware. On top of that, it’s natively multimodal, supporting image, text, audio, and video inputs 🤯

Today, Gemma 3n is finally available on the most used open source libraries. This includes transformers & timm, MLX, llama.cpp (text inputs), transformers.js, ollama, Google AI Edge, and others.

This post quickly goes through practical snippets to demonstrate how to use the model with these libraries, and how easy it is to fine-tune it for other domains.

Models released today

Here is the Gemma 3n Release Collection

Two model sizes have been released today, with two variants (base and instruct) each. The model names follow a non-standard nomenclature: they are called gemma-3n-E2B and gemma-3n-E4B. The E preceding the parameter count stands for Effective. Their actual parameter counts are 5B and 8B, respectively, but thanks to improvements in memory efficiency, they manage to only need 2B and 4B in VRAM (GPU memory).

These models, therefore, behave like 2B and 4B in terms of hardware support, but they punch over 2B/4B in terms of quality. The E2B model can run in as little as 2GB of GPU RAM, while E4B can run with just 3GB of GPU RAM.

Size	Base	Instruct
2B	google/gemma-3n-e2b	google/gemma-3n-e2b-it
4B	google/gemma-3n-e4b	google/gemma-3n-e4b-it

Details of the models

In addition to the language decoder, Gemma 3n uses an audio encoder and a vision encoder. We highlight their main features below, and describe how they have been added to transformers and timm, as they are the reference for other implementations.

Architecture Highlights

Performance & Benchmarks:

Demo Space

The easiest way to vibe check the model is with the dedicated Hugging Face Space for the model. You can try out different prompts, with different modalities, here.

📱 Space