Google has unveiled Gemma 4, its latest open-weight model family, built on research from Gemini 3. This family includes four models: the Effective 2B (E2B) and Effective 4B (E4B) edge models, optimized for devices like Raspberry Pi and smartphones, as well as the 26B Mixture-of-Experts (MoE) and 31B Dense models. The introduction of the Apache 2.0 license marks a significant shift, allowing wider commercial deployment and greater flexibility for developers. Demis Hassabis, CEO of Google DeepMind, emphasizes that these are “the best open models in the world” for their size. Enhanced capabilities include multi-step reasoning and faster performance, with the E2B model running three times faster than E4B. Gemma 4 supports multimedia processing in over 140 languages and highlights significant battery efficiency. With over 400 million downloads, Gemma’s ecosystem exemplifies robust developer adoption, now accessible on platforms like Hugging Face and Google AI Studio.
Source link
Share
Read more