Google Releases Gemma 3 — Here’s What You Need To Know

Image from Google

Google released a brand new model called Gemma 3 with 27 billion parameters. Gemma is a family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models. Gemma 3 is designed for developers building AI apps that can run anywhere from phones to workstations with support for over 35 languages and the ability to handle text, images, and short videos.

According to the company’s blog post, it’s the “world’s best single-accelerator model,” outpacing Meta’s Llama, DeepSeek, and OpenAI’s o1-preview and o3-mini-high on a single GPU host. That’s impressive for a model built to stay lean.

Here’s what you need to know about Gemma 3.

What is Gemma 3?

Unlike Google’s proprietary Gemini models, which power their consumer-facing tools, Gemma 3 is open-source. It means it’s accessible and available for use to anyone. It comes in four sizes: 1 billion, 4 billion, 12 billion, and 27 billion parameters.

The new model introduces several key features:

• Image and Text Input: Multimodal capabilities enable you to input both images and text, allowing for deeper analysis and understanding of visual data.
• 128K Token Context: Offers a 16x larger context…