Watch: Google launches ‘Gemma 3’ - Here's what it can do

American tech juggernaut Google has released Gemma 3 – the successor to their range of lightweight open models powered by their Gemini 2.0 models.

Gemma 3 is Google's latest open-source large language model (LLM), designed to understand and generate human-like text. It analyses text, images, and short videos, making it versatile for various applications. Notably, Gemma 3 is engineered to run efficiently on a single GPU, allowing developers to deploy advanced AI capabilities without requiring extensive hardware resources.

In a blog post, Google said Gemma 3 was available in sizes 1B, 4B, 12B and 27GB. The tech company claimed these models are their “most advanced, portable and responsibly developed open models yet.”

The models are designed to run fast, directly on devices — from phones and laptops to workstations — helping developers create AI applications. Last month, Google celebrated Gemma's first birthday, a milestone marked by over 100 million downloads.

The company claims that it’s the “world’s best single-accelerator model,” outperforming competition from Facebook’s Llama, DeepSeek, and OpenAI for performance on a host with a single GPU, as well as optimised for running on Nvidia’s GPUs and dedicated AI hardware.

Gemma 3’s vision encoder is also upgraded, supporting high-res and non-square images. In contrast, the new ShieldGemma 2 image safety classifier is available to filter both image input and output for content classified as explicit, dangerous, or violent.

Go global in 140 languages

Gemma 3 offers out-of-the-box support for over 35 languages and pre-trained support for over 140 languages. It can quickly build applications that analyse text and short videos, opening up new possibilities for interactive and intelligent applications.

It can also handle complex tasks with an expanded context window. “Gemma 3 offers a 128k-token context window to let your applications process and understand vast amounts of information,” it explained.