Distinguishing neurosarcoidosis (NS) from multiple sclerosis (MS) remains challenging and available parameters lack discriminatory power. Comprehensive flow cytometry data of blood and CSF leukocytes of patients with NS (n = 24), MS (n = 49) and idiopathic intracranial hypertension (IIH, n = 52) were analyzed by machine learning algorithms. NS featured a specific immune cell pattern with increased activated CD4+ T cells in CSF and increased plasma cells in blood. Combining blood and CSF parameters improved the differentiation. We thereby identify and independently validate a multi-dimensional model of blood and CSF supporting the difficult differential diagnosis between NS and MS.