Fugaku Supercomputer Powers Detailed Mouse Brain Simulation

Researchers have advanced brain simulation by modeling the mouse cortex with high detail, using the computational resources of the Fugaku supercomputer in Japan. This development demonstrates the use of large-scale computing to study complex brain activity at the level of individual neurons and synapses.

The project brought together biological data from the Allen Cell Types Database and Allen Connectivity Atlas, enabling the simulation to reflect detailed neuronal behavior. The simulation reproduced processes such as ion flows and changes in membrane voltage, allowing observation of spontaneous activity in the cortex.

Fugaku, recognized as one of the fastest supercomputers globally, provided the necessary processing power for this research. The simulation involved nearly 10 million neurons and 26 billion synapses distributed across 86 brain regions, representing a substantial computational challenge.

Researchers involved in the project stated that the current model does not include certain features such as brain plasticity, neuromodulators, or detailed sensory input. These limitations were noted as areas for future development in the simulation models.

What the numbers show

• Nearly 10 million neurons and 26 billion synapses were modeled
• Simulation spanned 86 interconnected regions of the mouse cortex
• One second of brain activity was simulated in no more than 32 seconds

Anton Arkhipov of the Allen Institute said in a statement that this achievement demonstrates the technical feasibility of even larger brain simulations. According to the researchers, future efforts may extend to modeling entire brains, potentially including those of monkeys or humans.

The simulation was presented by the University of Electro-Communications and Allen Institute at the SC25 supercomputing conference in November 2025. This presentation highlighted the collaboration between institutions and the use of advanced computing infrastructure for neuroscience research.

The long-term objective, as stated by the research team, is to expand simulations beyond the mouse cortex. The project aims to develop models that could eventually represent whole-brain activity in more complex species, using similar computational approaches.

By combining large-scale biological datasets with high-performance computing, the research offers a framework for future studies of brain function. The current results provide a basis for further refinement and expansion of digital brain models.

* This article is based on publicly available information at the time of writing.