China claims brain-like AI breakthrough ‘100 times faster than traditional models’
Chinese researchers from the Chinese Academy of Sciences have announced a major advancement with SpikingBrain 1.0, a “brain-like” AI model that emulates the way human neurons selectively activate. Unlike conventional large language models that analyze entire sentences word by word, SpikingBrain mimics human cognition by focusing on the most relevant nearby words. This selective processing dramatically improves efficiency and speed—researchers claim performance up to 25 to 100 times faster than traditional AI systems.
Another key advantage is SpikingBrain's economy of scale in both data and infrastructure. It can achieve comparable performance with less than 2% of the training data, and it operates on China’s domestically developed MetaX chip platform rather than relying on Nvidia hardware—demonstrating greater independence amid global tech restrictions.
By combining brain-inspired algorithms with local chip architecture, SpikingBrain marks a strategic step in developing AI technologies that are both energy-efficient and geopolitically resilient. While the study has yet to undergo peer review, it signals China's growing capabilities in neuromorphic computing—AI designed to mirror the human brain—and its intention to reduce reliance on Western technology.