Google has introduced a set of three new generative AI models, described as “open,” “safer,” “smaller” and “more transparent” compared to most others — a claim that could be seen as ambitious.

These models are additions to Google’s Gemma 2 family of generative models, which was first introduced in May. The newly released models, Gemma 2 2B, ShieldGemma, and Gemma Scope, are created for somewhat different applications and scenarios, but all share a focus on safety.

The Gemma series of models from Google differ from its Gemini models in that Google does not release the source code for Gemini, which is utilized in Google’s own products and also made accessible to developers. Instead, Gemma represents Google’s attempt to build positive relationships within the developer community, similar to Meta’s efforts with Llama.

Gemma 2 2B is a compact model for generating and analyzing text that can operate on various hardware types, including laptops and edge devices. It is licensed for certain research and commercial uses and can be obtained from platforms such as Google’s Vertex AI model library, the data science platform Kaggle, and Google’s AI Studio toolkit.

Regarding ShieldGemma, it’s a group of “safety classifiers” that try to identify toxicity such as hate speech, harassment and sexually explicit material. Developed on top of Gemma 2, ShieldGemma can be utilized to screen prompts to a generative model as well as content that the model creates.

Finally, Gemma Scope enables developers to “zoom in” on particular points within a Gemma 2 model and render its internal processes more comprehensible. Google describes it in a blog post as follows: “[Gemma Scope is made up of] specialized neural networks that help us unpack the dense, complex information processed by Gemma 2, expanding it into a form that’s easier to analyze and understand. By studying these expanded views, researchers can gain valuable insights into how Gemma 2 identifies patterns, processes information and ultimately makes predictions.”

The introduction of the new Gemma 2 models follows shortly after the U.S. Commerce Department endorsed open AI models in a preliminary report. The report indicated that open models extend the accessibility of generative AI to smaller companies, researchers, nonprofits, and individual developers while also stressing the necessity for capabilities to monitor such models for potential risks.