Google Releases Gemma 4 Open AI Models

Google has introduced Gemma 4, its most intelligent open models to date, designed for advanced reasoning and agentic workflows. Built with technology from Gemini 3, these models offer high performance-per-parameter and are available under an Apache 2.0 license. Released in four sizes, including 2B, 4B, 26B MoE, and 31B Dense, Gemma 4 excels in complex logic, function-calling, and multimodal tasks. The models are optimized for diverse hardware, from mobile devices to workstations, enabling efficient fine-tuning and local AI applications.