Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

Google releases Gemma, open models

·

Google publishes the weights of Gemma 2B and 7B, built on the research behind Gemini.

A worker views two purple model blocks on a laptop beside a local server.
Gemma brought small models you can run yourself. Illustration: Denkbaar.

On 21 February 2024, Google releases Gemma: a family of small open models.

What happened

  • Google publishes the model weights in two sizes: Gemma 2B and Gemma 7B.
  • Each size has a pre-trained version and an instruction-tuned version.
  • Google says Gemma comes from the same research as the Gemini models. The name comes from the Latin gemma, precious stone.
  • There is also a toolkit that helps you build safer applications.
Two small and two larger model blocks lie in pairs beside each other.
Each Gemma size had a base version and an instruction version. Illustration: Denkbaar.

Why it mattered

With this, Google joins the labs whose model weights you can download yourself. You can run such models on your own machine or server.

A worker reads a licence beside a laptop with a purple model block.
The licence decides what you may do with an open model in business. Illustration: Denkbaar.

What you can do with it now

“Open model” does not tell you everything. Always read the licence: it decides what you may do with the model in your business.

Written afterwards on 12 November 2026, from the source below.

Sources

  1. Google: Gemma: Google introduces new state-of-the-art open models

Topics: open-sourcemodelsgoogle