Nine Million Biophysical Neurons, the Mouse Cortex on Fugaku
Rin Kuriyama, Tadashi Yamazaki, Anton Arkhipov and colleagues published Microscopic-Level Mouse Whole Cortex Simulation Composed of 9 Million Biophysical Neurons and 26 Billion Synapses on the Supercomputer Fugaku in the proceedings of SC ‘25, the ACM supercomputing conference, on November 15, 2025 (DOI 10.1145/3712285.3759819). The paper reports a scale record with a detail level attached. Not nine million point neurons, nine million biophysical ones, covering a whole mouse cerebral cortex with 26 billion synapses.
| Measure | Value |
|---|---|
| Neurons simulated | 9 million, biophysical |
| Synapses | 26 billion |
| Strong-scaling benchmark | 152,064 compute nodes on Fugaku |
| Sustained performance | 7.13 petaflops |
What biophysical detail costs, and what it buys
The paper’s method section carries as much weight as its record. The authors built a lightweight biophysical neuron simulator and optimized it for Fugaku’s architecture. The strong-scaling benchmark reached 152,064 compute nodes while sustaining 7.13 petaflops, and the realistic scenario then simulated the whole mouse cerebral cortex at microscopic level. Biophysical here is the operative word. Unlike the leaky integrate-and-fire models that dominate large-scale simulation, examined in the substrate console’s documentation, biophysical neuron models carry membrane dynamics, ion channels and compartmental structure, which is the detail tier the 2008 whole brain emulation roadmap placed several levels above simple integrate-and-fire, examined in the roadmap review.
The scaling numbers quantify the road to human scale. A mouse cortex at 9 million biophysical neurons took a petascale-to-exascale machine running at near its full node count. The human cortex carries roughly three orders of magnitude more neurons, and the computational cost of biophysical detail grows with both neuron count and per-neuron complexity. The paper does not promise the human brain. It prices the first unit of the climb.
Where this sits in the emulation record
The site’s emulation coverage now has three scale markers to compare. The Eon fruit fly demonstration, examined in the fly emulation analysis, paired a complete connectome-derived model with an embodied body and behavioral validation at 95 percent on motor tasks. The FlyWire and MaleCNS datasets, examined in the FlyWire analysis and the MaleCNS comparison, supply the wiring at synaptic resolution. The Fugaku result supplies biophysical runtime at rodent cortical scale, browsable alongside the wiring datasets in the connectome atlas. Wiring, dynamics and embodiment are the three legs of the emulation program, and each has now been demonstrated separately at a different scale.
Comparison to The Consciousness AI
This project treats consciousness as an emergent property of organized physical activity, and the Fugaku result is feasibility evidence for the physical-activity half of that claim at a new scale. The project’s research code, maintained in the tlcdv/the_consciousness_ai repository, keeps the claim scoped the way the paper keeps its own. A biophysical simulation of a cortex is a model of the dynamics. Whether those dynamics carry experience, in silicon or in silico, is the question the simulation was not designed to answer, and this site’s discipline is to neither smuggle the answer in nor deny it in advance.
What the Record Leaves Open
The paper establishes that biophysical whole-cortex simulation at mouse scale fits on existing supercomputing hardware. It leaves open the three gaps standing before the emulation program, the wiring data at mouse-cortex scale is incomplete, the validation against behavior is not yet demonstrated at this detail tier, and the human-scale extrapolation remains bounded by memory and energy rather than by principle. Those gaps are the working agenda of the brain emulation page, and the field’s state of the argument is reviewed in the flagship overview of AI consciousness research.
Researchers covered here
- Rin KuriyamaFirst author of the SC '25 whole mouse cortex simulation on Fugaku
- Tadashi YamazakiUniversity of ToyamaSpiking neural network simulations on Japanese supercomputers, senior author of the SC '25 Fugaku mouse cortex result
- Anton ArkhipovAllen Institute for Brain ScienceLarge-scale neural simulation, biophysically detailed cortical models, AIBRAIN program, State of Brain Emulation Report 2025