Google has not taken its foot off the pedal when it comes to AI model announcements. Fresh off revealing Google Gemini for Workspace, the company has unveiled Gemma, its free, open-weight AI model. Unlike Gemini, Gemma is available to anyone for no cost, though it is less powerful than its subscription-based sibling.
WATCH: OpenAI’s Sora creates photorealistic videos from text prompts
Google’s new offering is available in two sizes, 2B (2 billion parameters) and 7B (7 billion parameters), each of which has pre-trained and instruction-tuned variants to choose from. Besides being free, Gemma has its own share of advantages. Sacrificing sheer power for speed and convenience, Gemma is lightweight and can be run locally on almost any computer. This positions Gemma to be better suited for light tasks such as simple summarizations or as a chatbot.
Gemma is being released with alongside a Responsible Generative AI Toolkit, which Google says “provides guidance and essential tools for creating safer AI applications with Gemma.” Though Gemma is strictly less powerful than Gemini, Google claims that its testing shows that Gemma punches above its weight class, outperforming larger models such as Meta’s Llama 2 comparable 7B and 13B models in metrics such as reasoning, math, and coding. This may have something to do with Google’s assertion that Gemma was built from the same research and technology used to create Gemini. Gemma is available now to try for free.
