OBJECTIVE: Cognitive deficits present transdiagnostic characteristic and partly explain the poor functional outcomes of patients with mental disorders. Understanding the relationships between neurocognition, social cognition, and global function may help identify new cognitive intervention targets. We aimed to model the complex interrelationships among these variables with Gaussian Graphical Modeling in a transdiagnostic sample. METHODS: A total of 482 individuals were included in this study, comprising 281 patients with first-episode schizophrenia, 128 patients with bipolar disorder, and 73 patients with major depressive disorder. The Wechsler Adult Intelligence Scale, the MATRICS Consensus Cognitive Battery, and the Global Assessment of Functioning Scale were evaluated. The interaction and centrality indexes of cognitive and global function were analyzed by network analysis. RESULTS: In the transdiagnostic network, speed of processing (SOP) and verbal learning (Vrbl) exhibited higher centrality indexes. The cognitive nodes closely associated with global function included working memory (WM), and attention/vigilance (AV). When subjects were modeled separately by gender, no significant differences were found between males and females. CONCLUSION: The close connections between WM, AV, and global function as well as the high centrality indexes of SOP and Vrbl suggest that these domains share aspects of pathophysiology in schizophrenia and mood disorder. However, the data-driven approach limited our interpretation of the results. Theory-driven model should be further validated to elucidate causal pathways and find more promising approaches to recovery.