Google has unveiled its latest advancement, Gemma 4, an open model optimized for advanced reasoning and agentic workflows. Available in four sizes—including Effective 2B (E2B) and Effective 4B (E4B)—Gemma 4 outperforms models 20 times its size, achieving top rankings on the Arena AI text leaderboard. This model supports over 140 languages and excels at multi-step logic, code generation, and handling complex inputs with context windows up to 256K tokens. Additionally, users can fine-tune Gemma 4 on various platforms, including Nvidia and AMD hardware. Meanwhile, Google Vids introduces new features such as AI-generated videos with Veo 3.1, allowing users 10 free video generations monthly and enhanced capabilities for Google AI Pro subscribers. The new Vids Screen Recorder and YouTube integration further streamline content creation, enabling seamless video publishing. Released under the Apache 2.0 license, Gemma 4 is available on Hugging Face and Google AI Studio.
