Abstract Background Although various prognostic models for primary central nervous system lymphoma (PCNSL) have been developed, there is no consensus regarding the optimal prognostic index. We aimed to evaluate potential prognostic factors and construct a novel predictive model for PCNSL patients. Methods We enrolled newly diagnosed PCNSL patients between 2003 and 2015. The primary endpoint was progression‐free survival (PFS), and the secondary endpoint was overall survival (OS). The prognostic factors identified using multivariate Cox proportional hazards models were used to develop a predictive model. We subsequently validated the prognostic model in an independent cohort. We also evaluated the validity of the existing scores: the International Extranodal Lymphoma Study Group (IELSG), the Nottingham/Barcelona (NB), and the Memorial Sloan‐Kettering Cancer Center models (MSKCC). Results We identified 101 patients with newly diagnosed PCNSL at our center. Multivariate analysis showed that age ≥80, deep brain lesions, and ECOG ≥2 were independent risk factors of PFS. Assigning one point for each factor, we constructed a novel prognostic model, the Taipei Score, with four distinct risk groups (0‐3 points). The performances of the Taipei Score in discriminating both PFS and OS in the training cohort were significant, and the score was validated in the external validation cohort. The IELSG, NB and MSKCC models had insufficient discriminative ability for either PFS or OS in both cohorts. Conclusion The Taipei Score is a simple model that discriminates PFS and OS for PCNSL patients. The score may offer disease risk stratification and facilitate clinical decision‐making.