Following the recent announcement of the new Gemini model, Google is now launching a family of lightweight “open” models: Gemma 2B and Gemma 7B.

Google is far from finished with its AI developments. Last week Google launched the latest version of its Gemini models, now the company is knocking on the door with the launch of Gemma, a family of lightweight, open models with two versions: Gemma 2B and Gemma 7B. The model was developed using the same research and infrastructure elements used to create the Gemini models. Like the recently launched Gemini 1.5 Pro, the Gemma models are also available in Vertex AI. Although it is described as an “open” model, we should not confuse it with an open source model.

“Open” model

Google’s new Gemma 2B and Gemma 7B models are based on the same technology used to create the Gemini models. According to Google, these models are said to be very powerful for their size and perform better compared to comparable models from Meta or Mistral, for example. Developers get access to pre-built Colab and Kaggle notebooks, as well as integrations with Hugging Face, MaxText, and Nvidia’s NeMo. The models contain pre-trained and tuned variants that can be used on a laptop, workstation or in the Google Cloud.

Gemma, Latin for gemstone, is an “open” model, but not an open source model. This is an important distinction. With open source models, users have complete creative autonomy without restrictions.

Finally, the new Gemma models are equipped with Google’s new ones Responsible generative AIToolkit. With this, Google wants to support its users with best practice resources to generate safer AI applications and to provide debugging tools.