Google’s Gemma 3n Is Now Backed by Filecoin for Storage

• Google DeepMind’s Gemma 3n runs locally with multimodal support and low memory usage using MatFormer and Per‑Layer Embeddings.
• Eternal AI preserved Gemma 3n on Filecoin, enabling decentralized access and long-term verifiable storage via CLI or desktop.

Google DeepMind is redrawing the map of artificial intelligence with the launch of Gemma 3n, a lightweight multimodal model that can run on local devices. Not only high-end laptops, this model is also designed to remain agile on mobile phones and computers with limited memory.

Interestingly, Eternal AI decided to store this model on the decentralized Filecoin network. This step is not only about backup, but also about ensuring that Gemma 3n remains publicly accessible—without relying on traditional cloud services.

Gemma 3n comes in two versions, each with 5 and 8 billion parameters. Although it sounds heavy, this model only takes up around 2 to 3 GB of memory thanks to approaches such as Per-Layer Embeddings and MatFormer. You could say, this is like having a supercomputer in the palm of your hand. This model also supports text, image, voice, and video input.

Gemma 3n, the new multimodal model by Google DeepMind, is now preserved on Filecoin by @CryptoEternalAI. It supports text, image, audio, and video, runs on local devices, and can be stored, verified, and retrieved without centralized services via CLI or desktop. https://t.co/Vo5E9DydWB

— Filecoin (@Filecoin) June 28, 2025

Filecoin Powers a New Standard for Verifiable, Permanent Data

On the other hand, the decision to store Gemma 3n on Filecoin is not a whim. Eternal AI chose Filecoin because this protocol guarantees security, transparency, and data availability. With a cryptographically verifiable CID (Content Identifier) ​​system, anyone can be sure that the file they are taking is the original file—without having to ask any data center first.

On the other hand, CNF reported a few days ago that Filecoin has also successfully stored the entire history of the Solana blockchain in a decentralized storage system. This means that developers and AI agents now have access to clean, verified, and tamper-proof data.

This is not just about efficiency, but also about trust. Because as we know, data that can be secretly changed can make AI go haywire.

However, this is not the only role of Filecoin in the data preservation mission. The Flickr Foundation, for example, has uploaded the 1,000 most viewed cultural photos to the Filecoin network. The reason? They want to keep this archive available to the public, even if the original server shuts down.

Meanwhile, DocumentCloud has also used Filecoin to secure court documents and government archives from the risk of being lost or forcibly deleted.

Meanwhile, as of press time, FIL is changing hands at about $2.31, up 2.22% over the last 24 hours and 10.48% over the last 7 days.

Gemma 3n and the Rise of Self-Sovereign AI Infrastructure

The combination of local inference and decentralized storage may sound technical, but it’s really like having an electric car that can charge itself from its own solar panels. It’s economical, self-sufficient, and there’s no need to wait in line. With a model like Gemma 3n, users can run AI without having to send personal data to the cloud. And with Filecoin, those files stay safe even if the main server goes offline.

For developers, this opens up a world of possibilities. Imagine building a medical or legal app that requires sensitive data but can’t leak it. Or an AI model that can run on your phone offline, but can still be updated from the Filecoin repository when a connection is available.

In other words, Gemma 3n isn’t just a new technology, it’s part of an ecosystem that’s trying to change the way we think about AI—making it more self-sufficient, more tamper-proof, and less likely to be tracked by third parties.

It also shows that the future of AI doesn’t have to be tied to big companies or servers that are in the middle of nowhere. It could be that more home developers, creators, or small communities are the ones who will be the pioneers of the next generation of AI.