Google releases Gemma 4

Google released Gemma 4, an open-source model in four sizes (2B–31B) with up to 256K context and AI agent support. Runs on consumer hardware without a data center.

Author: Michael Kokin ·

A lightweight open-source model that nearly matches the best closed-source AI. And it runs on consumer hardware.

What's inside:
• 4 sizes: from a tiny 2B to a powerful 31B Dense
• 128K–256K context window
• Built for real-world tasks and AI agents

The only top-3 open model that doesn't require a data center. A laptop or a local server is enough — you can even run agents on it.

Available now on Hugging Face. Free for commercial use.