Google releases open source model Gemma 4

On April 2nd local time, Google announced the launch of its new generation open-source model, Gemma 4, claiming it to be the "smartest" open-source model to date, mainly targeting advanced reasoning and intelligent agent workflow applications. Google stated that Gemma 4 is built on the same research and technology system as the "Gemini 3" model, with fewer restrictions on commercial use, making it the most powerful open-source model in terms of capabilities at the same scale. Gemma 4 was released in four specifications to adapt to different hardware environments ranging from mobile devices, laptops, development workstations, and accelerators. Larger models can support higher levels of offline inference on personal computers, while smaller models are focused on low-latency, multimodal, and offline capabilities on mobile and IoT devices.