Google unveils Gemma 2 models with enhanced safety, interpretability, and performance

Google has just announced three new additions to the Gemma 2 family, Google's next-generation of open large language models (LLMs). The lineup includes:

1. Gemma 2 2B – a new iteration of Google's 2 billion (2B) parameter model, now featuring built-in safety advancements and optimized for a balance of performance and efficiency.
2. ShieldGemma – a suite of safety content classifier models based on Gemma 2, designed to filter the input and outputs of AI models to ensure user safety.
3. Gemma Scope – a model interpretability tool providing unparalleled insight into the inner workings of Google models.

With these additions, researchers and developers can leverage Google Gemma to create safer customer experiences, gain deep insights into model operations, and confidently deploy advanced AI on devices, paving the way for new innovations.