HQ500M
Virtual neuron addressing, sparse CPU dynamics, dendritic proxies, plasticity and verifiable checkpoints.
We build experimental neural architectures at extraordinary scales, then test what actually happens. Anatomical structure. Plasticity. Computation. Evidence over mythology.
A neural system is more than a number. Our work spans addressable populations, active dynamics, structure-aware memory, and the cost of putting them together.
Different milestones have different levels of evidence. We show those differences instead of erasing them.
Virtual neuron addressing, sparse CPU dynamics, dendritic proxies, plasticity and verifiable checkpoints.
Scaling the synthetic architecture and exploring more demanding compute and memory strategies.
A numerical scale target, not proof of a human mind, complete brain connectivity or biological equivalence.
✳ Numbers describe virtual model populations or research targets, not measured biological neurons simultaneously simulated.
Explore 283 individually named anatomical structures from the Human Reference Atlas. Rotate the brain, reveal the cerebellum, select regions and inspect their UBERON identities. Our synthetic HQ500M connections remain in a separate research view.
NEUTRALLAB ↗VIEWPORT / 001HRA ATLAS · INTERACTIVE◉A biological wiring diagram, a synthetic model and a hypothesis are three different things. NeutralLab keeps the boundaries visible.
READ THE SOURCE ↗Only actual source data and reproducible test outcomes qualify. Every benchmark belongs to its hardware, date and method.
Virtual neurons, algorithmic wiring and procedural 3D anatomy are synthetic abstractions with explicit limitations.
Neural counts alone do not establish cognition, consciousness or anatomical fidelity. Claims require independent evidence.
Measured in a bounded synthetic HQ500M CPU workload. Only 32,768 of those IDs spiked. This is not a million simultaneous firing neurons and not an artificial human brain.
SOURCE & METHOD ↗Comparison within one documented CPU test. Not an apples-to-apples benchmark of biological activity.
Source code, benchmark artifacts, architectural notes and experimental limits. The research record should outlive any screenshot.