Post by Google

Today we're introducing Gemma 4 — our newest family of open models. Built from the same world-class research as Gemini 3, this family of models brings breakthrough intelligence directly to your own hardware. From autonomous agents that can plan and navigate apps to local-first coding assistants, Gemma 4 is designed for the most complex reasoning tasks. Here’s what’s new: 🔹 Longer context — up to a 256K context window to seamlessly process large datasets. 🔹 Four versatile sizes — tailored for everything from mobile phones to developer workstations. 🔹 Native multimodality — built-in support for vision and audio processing. 🔹 Global reach in 140+ languages — natively trained on over 140 languages to help you build for a worldwide audience. 🔹 Apache 2.0 license — a commercially permissive license that gives you total control to build freely and deploy securely. By bringing our most advanced proprietary research into an open ecosystem, we’re giving you everything you need to build the next generation of AI — anywhere you want. You can start building with Gemma 4 now in Google AI Studio, or download the model weights from Hugging Face, Kaggle, and Ollama. Learn more → goo.gle/4s9G1yQ

Video Content