Google unveiled its latest series of open-source large language models, dubbed Gemma, on Wednesday, offering them for free. Gemma is constructed from the same research foundation as one of Google’s existing models, Gemini, and is available in two variants: Gemma 2B and Gemma 7B. These versions are compatible with Google Cloud, workstations, and developers’ laptops.
The tech giant announced that first-time Cloud users will receive a $300 credit, highlighting the optimization of this model for Google Cloud. Gemma is a joint development effort between Google DeepMind and the Google AI team. According to Google, the name “Gemma” is derived from the Latin word ‘Gem,’ signifying a “precious stone.”
In order to provide support and guidance to developers, Google also released the “Responsible Generative AI Toolkit,” aimed at promoting the safe and responsible development of AI applications. It’s worth noting that Google retains the right to modify the terms and conditions of use, given that this open-source project is still in development. Gemma offers functionalities beyond typical language models, such as chatbot capabilities and content generation.
The primary objective behind Gemma is to democratize artificial intelligence and ensure its safe and responsible usage through the accompanying toolkit. Designed to operate efficiently even in low-resource environments, Gemma boasts a vocabulary of 250,000 tokens and can tackle tasks ranging from solving math problems to coding. This enhances the accuracy and relevance of the data it provides.
Chip manufacturer NVIDIA announced its collaboration with Google to ensure seamless integration of Gemma models with its chips. Additionally, NVIDIA plans to develop chatbot software compatible with Gemma for Windows PCs.
With the launch of Gemma, Google is following a strategy reminiscent of Meta platforms, aiming to entice more developers to its Vertex AI Cloud platform.