Understanding how cortical circuits generate complex behavior requires investigating the cell types that comprise them. Here, much effort has been focused on inhibitory neuron types but the functional roles of distinct classes of excitatory pyramidal neurons (PyNs) are less well understood. We, therefore, used widefield imaging to measure the cortex-wide activity of distinct PyN types and investigated their functional role in mice that performed an auditory decision-making task. We used two mouse lines, expressing the calcium indicator GCaMP6s in two major PyN types: FezF2 for pyramidal-tract (PT) and PlexinD1 for intratelencephalic (IT) neurons. Using dimensionality-reduction methods, we isolated cortex-wide activity patterns of PT and IT neurons and compared them to EMX mice with GCaMP6s-expression in all PyNs. We found major PyN-specific differences in the complexity and spatial layout of cortical activity patterns, both at the local and mesoscale, suggesting the existence of specialized subcircuits. We also found PyN-specific functional differences during decision-making. Sensory responses were largest in the sensory, parietal, and frontal cortex but each PyN type showed pronounced differences in cortical localization and spatial specificity. The same was true for choice-related activity: A choice decoder revealed ramping, and contralateral choice-selective activity in parts of the frontal cortex of EMX and PT mice whereas IT mice showed ipsilateral choice signals. Using an inter-sectional viral strategy, we found that this inverse choice tuning in IT was most pronounced in corticostriatal projection (CStr) neurons. Lastly, we used optogenetic inhibition to causally test the importance of PyN-types for decision-making. Inactivating the parietal cortex disrupted sensory processing, with the strongest effect in PT neurons. In the frontal cortex, all PyN-types reduced animal performance, suggesting that they are equally involved in choice formation and execution. Our work reveals PyN-specific, cortex-wide dynamics and strongly supports the view that local circuits throughout the cortex perform parallel computations, even within the same cortical layer.