Skip to content

AI Made

AI agents, automation, and tech journalism

Google Just Made Gemma 4 Completely Open — And It Runs on a Raspberry Pi

Hey guys, Mr. Technology here. I’ve been waiting for this moment for a long time — a genuinely capable AI model that you can run on your own hardware, with no API calls, no data leaving your building, and no commercial restrictions.

What You Need to Know:

  • Google dropped Gemma 4 with full Apache 2.0 licensing — no restrictions, no royalty fees, full commercial rights
  • The 7B model runs on a Raspberry Pi 5 at approximately 18 tokens per second (INT4)
  • Consumer GPU fine-tuning takes under 2 hours with Google’s provided toolkit
  • Medical, legal, and financial teams are already building specialized local models with it

I went hands-on with the full Gemma 4 release in my review of Google Gemma 4 running on Raspberry Pi and whether it’s actually practical for real-world tasks.

## Why Apache 2.0 Matters More Than You Think

This isn’t just a license technicality. Full Apache 2.0 means medical companies can build diagnostic aids without worrying about usage restrictions. Law firms can fine-tune on client data without data leaving their network. Schools can run models without internet access.

The practical applications of “truly open” go way beyond the developer community.

## The Fine-Tuning Story

Two hours on a consumer GPU and you have a specialized model trained on your proprietary data. No API calls. No data sharing. No vendor dependency.

That’s not a toy. That’s a fundamental shift in what’s possible for organizations that couldn’t previously afford AI infrastructure.

What do you think? Is Gemma 4 the real deal or overhyped? Drop your thoughts below!