Chinese researchers have revealed a novel neuromorphic computer modeled on a monkey’s brain, named “Darwin Monkey”—also referred to as “Wukong,” after the mythological Chinese figure who inspired characters like Goku from Dragon Ball and other pop culture icons.
Development took place in the National Key Laboratory of Brain-Computer Intelligence at Zhejiang University, achieving a new benchmark in efficiency and parallel computing. Darwin Monkey contains 2 billion spiking neurons and more than 100 billion synapses, similar to the brain of a type of monkey called a macaque.
Previous Work in Neuromorphic Computing
Neuromorphic computing, a branch of computer science focused on mimicking the brain’s architecture and functions, has seen rapid advancements in recent years amid the global AI race. The new system represents a significant leap beyond Zhejiang University’s 2020 project, “Darwin Mouse” (also known as “Mickey”), which featured 120 million artificial neurons and was modeled after a mouse brain. In just five years, the capabilities of neuromorphic systems have grown exponentially.
Until now, Intel’s 2024 Hala Point was considered the most advanced neuromorphic system, boasting 1.15 billion neurons. That system enabled large-scale data processing for highly complex AI tasks. Beyond its raw computing power, Intel emphasized how neuromorphic systems—by mimicking brain architecture—can drastically improve energy efficiency. These systems offer a compelling response to growing concerns about AI’s unsustainable consumption of water and power.
Designing Darwin
The Darwin 3 neuromorphic chip at the heart of the new system was developed in early 2023 through a collaboration between Zhejiang University and Zhejiang Lab. Each chip contains 2.35 million spiking neurons and synapses capable of operating at exascale speeds—up to 10¹⁸ (a billion billion) calculations per second. The chips also feature brain-inspired instruction sets and neuromorphic online learning systems.
Darwin Monkey comprises 15 servers, each hosting 64 of the advanced chips. The 12-inch blade servers were designed to maximize energy efficiency and reduce inter-chip communication bottlenecks.
Among the project’s key breakthroughs were the development of a large-scale neural system interconnection and integration architecture, an adaptive time-step control method, a wafer-thin brain-inspired chip design, and a data-swapping strategy for a multi-level memory system.
Operating a Brain-Inspired Computer
To fully harness the system’s potential, the researchers also developed a new operating system tailored to the unique demands of neuromorphic computing. This OS is load-aware and supports scheduling algorithms that optimize task allocation by considering both communication bandwidth and task characteristics.
In testing, the team ran the DeepSeek large language model on Darwin Monkey, tasking it with content creation, mathematical problem solving, and logical reasoning.
While the system has not yet reached the capabilities of the human brain, researchers believe it offers a powerful tool for brain research and the development of artificial intelligence. The team successfully simulated the brains of Caenorhabditis elegans, zebrafish, mice, and macaques using Darwin Monkey. With its immense scale, parallelism, and power efficiency, the system is well-suited for advanced AI experimentation. Its unsupervised online learning capabilities are expected to be particularly beneficial for future AI research.
Ryan Whalen covers science and technology for The Debrief. He holds an MA in History and a Master of Library and Information Science with a certificate in Data Science. He can be contacted at ryan@thedebrief.org, and follow him on Twitter @mdntwvlf.