Google's Gemma 4 model goes fully open-source and unlocks powerful local AI - even on phones

Key Points:

• Google DeepMind has released Gemma 4, its latest open large language model (LLM), under the Apache 2.0 license, making it fully open-source and allowing free use, modification, and redistribution without restrictions.
• Gemma 4 supports local deployment on a wide range of devices, from powerful servers to smartphones and IoT devices, enabling offline AI use that enhances privacy, reduces latency, and lowers operational costs.
• The model suite includes four versions: two high-end models (26B and 31B parameters) optimized for servers, and two lightweight models (E2B and E4B) designed for mobile and edge devices, all supporting advanced reasoning, multi-step planning, vision, audio, and multi-language capabilities.
• Since its initial release, Gemma has been downloaded over 400 million times with more than 100,000 variants created, and the open-source licensing of Gemma 4 is expected to accelerate adoption and integration into diverse applications.
• Key features of Gemma 4 include large context windows (up to 256K tokens), offline code generation, native processing of video and images, and support for over 140 languages, making it a versatile tool for developers and enterprises with data sovereignty needs.