Google has released Gemma 4, its most capable open model family to date, available in four sizes: Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts, and 31B Dense. The 31B model ranks #3 and the 26B ranks #6 on the Arena AI open model leaderboard, outperforming models 20x their size. Key capabilities include advanced reasoning, native agentic workflow support (function-calling, structured JSON), code generation, multimodal vision and audio input, context windows up to 256K tokens, and support for 140+ languages. All models are released under an Apache 2.0 license. The E2B and E4B edge models are optimized for mobile and IoT devices, running fully offline on Android phones, Raspberry Pi, and NVIDIA Jetson hardware. Day-one support is available across Hugging Face, Ollama, vLLM, llama.cpp, LM Studio, NVIDIA NIM, and many other platforms.
Sort: