Objective: The immunological features between neuromyelitis optica spectrum disorder (NMOSD), multiple sclerosis (MS), and myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD), lacked systemic comparisons. Accordingly, we aimed to investigate immunological differences between NMOSD, MS, and MOGAD. Methods: Patients with MOGAD, MS, and NMOSD who received immunological tests including cytokine profiles and cytometry analysis of the lymphocyte subgroups were retrospectively reviewed and divided into training and validation sets. Discriminatory models based on immunological data were established to identify optimal classifiers using orthogonal partial least square discriminant analysis (OPLS-DA). Constructed models were tested in another independent cohort. Results: OPLS-DA of the immunological data from 50 patients (26 NMOSD, 14 MS, and 10 MOGAD) demonstrated the discriminatory values of a relatively low level of T-lymphocyte subsets, especially the CD4+ T cells, in MOGAD; a decreased NK cell, eosinophil, and lymphocyte level; an elevated neutrophil-to-lymphocyte ratio in NMOSD; and a declined IFN-gamma-producing CD4+ T cells/Th with an increased IL-8 concentration in MS. All the models (NMOSD vs. MS, NMOSD vs. MOGAD, and MS vs. MOGAD) exhibited a significant predictive value and accuracy (>85%). Conclusions: NMOSD, MS, and MOGAD may be different in pathogenesis, and several immunological biomarkers can serve as potential classifiers clinically.