Google releases Gemma 4, a family of open models built off of Gemini 3

Google has launched Gemma 4, a new family of open-source AI models based on technology from its Gemini 3 system. The release includes four variants ranging from 2-billion to 31-billion parameters, designed for different devices from smartphones to powerful computers, with the larger models ranking third and sixth on Arena AI's text leaderboard despite being 20 times smaller than competitors. All models can process images and video, while smaller versions handle audio inputs, and they support over 140 languages with offline coding capabilities.