Google launches Gemma's new open-access language AI models

Google has announced the launch of Gemma, a new family of compact, open-access artificial intelligence language models. The first models introduced were Gemma 2B and Gemma 7B.

Here's What We Know

Google did not provide a detailed report on how these models perform compared to similar models from Meta and Mistral. The company only noted that they are "state-of-the-art."

According to Google's statement, Gemma is based on the architecture of the Gemini and PaLM models. It provides text generation capabilities comparable to the larger models, but in a compact size. This will allow developers to use it even on personal computers.

Although Gemma is called an open source model, its source code is not published. Users can customise the model and use it for inference, but cannot change the architecture itself.

To work with Gemma, Google provides out-of-the-box solutions based on Colab, Kaggle, Hugging Face, and other platforms.

The company has also released a set of tools for building responsible AI applications based on the new model.

Source: TechCrunch