Google has unveiled a new family of lightweight open-source AI models called Gemma, derived from the same technical framework as Gemini. This new AI model comes in two versions: Gemma 2B with 2 billion parameters and 7B with 7 billion parameters.Â
Google’s New AI Model Gemma
The previously mentioned models offer pre-trained and instruction-tuned variants optimized for various AI hardware platforms, including NVIDIA GPUs and Google Cloud TPUs.
However, despite its lightweight nature, it can run on developer laptops and desktops while delivering impressive performance.Â
Google claims this new lightweight AI model surpasses open models like Meta’s Llama 2 and outperforms larger models on key benchmarks. This new AI model prioritizes safety and reliability, using automated techniques to filter sensitive data during training.Â
Additionally, Google has developed a Responsible Generative AI Toolkit to ensure ethical behavior. More variants are expected, and developers can explore this new model on Google’s website.
Started his freelancing adventure in 2018 and began doing freelance Audio Engineering work and then started freelance writing a few years later.
Currently he writes for Gadget Pilipinas and formerly wrote for Grit.PH.
He is also a musician, foody, gamer, and PC enthusiast.
