Meta has unveiled Muse Glimmer, a new 30 billion parameter AI model developed by its Superintelligence Labs that can operate on consumer-grade hardware thanks to compressed weights. According to The Decoder, the model requires less than 20 GB of memory, making it accessible beyond high-end data centers.
Mark Zuckerberg has publicly defended the practice of distilling models from other leading AI labs and has called for fewer regulatory restrictions on US AI research. The Wall Street Journal reports that an open-weight version of Meta’s previous model, Muse Spark 1.2, is expected to be released soon.
This development is notable for Japanese investors and tech companies as it signals increased democratization of large AI models, potentially accelerating AI adoption in Japan’s competitive tech and financial sectors.