In mouse visual cortex experiments, Neuropixels Ultra increased neuronal yield by more than twofold. Its denser sampling captured the spatial footprint of extracellular electrical fields, helping distinguish axonal from somatic signals. The researchers also classified three genetically identified cortical interneuron types at about 80% accuracy and separated them from other neurons at about 85% accuracy.
The added density comes with a shorter recording span than Neuropixels 1.0, and the evidence is largely from animal experiments. The paper did not report simultaneous recording from 10,000 neurons. The 6,144 figure refers to selectable recording sites, which must be kept distinct from neuron count and simultaneous channels.