Subcortical brain structures are the hubs for various psycho-behavioral functions. There is no mega-analysis to simultaneously investigate subcortical volumetric alterations in schizophrenia, bipolar disorder, major depressive disorder, and autism spectrum disorder. Nor are there any neuroimaging data-driven clinical criteria overcoming limitations of the current diagnostic system, which would reflect cognitive/social functioning. We conducted a large-scale multisite study of subcortical volumetric and lateralization alterations in these disorders using T1-weighted images of 5,604 subjects (3,078 controls and 2,526 patients). We found schizophrenia-specific and cross-disorder shared alterations. Moreover, we clustered the 5,604 subjects based on subcortical volumes, and explored whether data-driven clustering results can explain cognitive/social functioning in the subcohorts. We showed a four-biotype classification, namely extremely and moderately smaller limbic regions, larger basal ganglia, and normal volumes, for predicting cognitive/social functioning. Our results will contribute to the future creation of novel biological data-driven psychiatry diagnostic criteria, expected to support appropriate treatment selection.