Gemma 3Gemma 3

Gemma 3: Advanced Open-Weight Language Model

Google DeepMind's latest open-weight large language model, built on Gemini 2.0 technology. Offering unprecedented performance with sizes from 1B to 27B parameters, supporting over 140 languages and a 128k-token context window.

Experience Gemma 3

Try out Gemma 3's advanced capabilities in text generation, visual reasoning, and multilingual support directly in your browser.

Experience the most capable model you can run on a single GPU or TPU.

What is Gemma 3

Gemma 3 is Google DeepMind's most advanced open model, designed to run efficiently on devices while delivering state-of-the-art performance.

Multiple Model Sizes

Available in 1B, 4B, 12B, and 27B parameter versions to suit different hardware capabilities and performance needs.

Extended Context Window

Process up to 128k tokens in a single context, ideal for handling large documents and complex tasks.

Open Source

Built for accessibility and innovation, allowing developers to customize and deploy on their preferred platforms.

How to Use Gemma 3

Get started with Gemma 3 in a few simple steps:

Key Features of Gemma 3

Explore the advanced capabilities that make Gemma 3 a powerful tool for AI development.

Function Calling

Built-in support for function calling and structured output to automate complex tasks.

Quantized Versions

Official quantized models for reduced computational requirements while maintaining accuracy.

Safety First

Comprehensive safety measures and ShieldGemma 2 integration for responsible AI development.

Hardware Optimized

Optimized for NVIDIA GPUs, Google Cloud TPUs, and AMD GPUs via ROCm™ stack.

Easy Integration

Seamless integration with popular frameworks including Hugging Face, PyTorch, and JAX.

Developer Tools

Comprehensive development support through Google AI Studio, Colab, and Vertex AI.

FAQ

Frequently Asked Questions

Find answers to common questions about Gemma 3.

What makes Gemma 3 unique?

Gemma 3 offers state-of-the-art performance while running on a single GPU/TPU, with advanced visual reasoning and multilingual capabilities.

What are the hardware requirements?

Gemma 3 is designed to run on various hardware configurations, from laptops to workstations, with different model sizes to suit your needs.

How can I access Gemma 3?

Access Gemma 3 through Google AI Studio, download from Hugging Face or Kaggle, or use the Google GenAI API.

What languages does Gemma 3 support?

Gemma 3 supports over 140 languages, with out-of-the-box support for 35+ languages.

Start Building with Gemma 3

Join the expanding Gemmaverse and create powerful AI applications.