Gemma 3 is a collection of lightweight, state-of-the-art open models built from the same research and technology that powers Google's Gemini 2.0 models. These models are designed to run fast on various devices, from phones and laptops to workstations, and are available in sizes ranging from 1B to 27B parameters. Gemma 3 supports over 140 languages, offers advanced text and visual reasoning capabilities, and includes a 128k-token context window and supports function calling for complex tasks. ## Sources - [homepage](https://ai.google.dev/gemma)