Google Launches Gemma, Open-Source AI Models Based on Gemini Tech
Google introduces Gemma, a lightweight open-source AI model family built on Gemini research. The models allow developers to run capable AI directly on laptops without needing expensive cloud access.
Google releases Gemma today, a new family of lightweight open-source AI models built using the same research and technology as its powerful Gemini model. Available in two sizes with two billion and seven billion parameters, Gemma provides developers and researchers with capable AI that runs directly on standard laptops and desktop computers.
Unlike the fully multimodal Gemini, Gemma focuses solely on text inputs and outputs in English. However, Google reports that these smaller models achieve best-in-class performance compared to similar open-source options, making advanced AI more accessible to those who cannot afford access to massive proprietary systems.
The tech giant designs Gemma with strict AI safety principles in mind, utilizing automated filtering to remove sensitive data and reinforcement learning from human feedback to encourage responsible behavior. By open-sourcing these models, Google gives the developer community full access to the underlying architecture so they can easily tune the technology to fit their specific needs.