Extranodal NK/T-cell lymphoma is an aggressive lymphoma that is strongly associated with Epstein-Barr virus infection. Although some extranodal NK/T-cell lymphoma patients have shown responses to immune checkpoint blockade, biomarkers for predicting extranodal NK/T-cell lymphoma patient response to immunotherapy have not yet been defined. To understand the tumor immune microenvironment, we analyzed the expression of 579 immune-related genes and characterized the immune cells using immunohistochemistries and in situ hybridization for EBER. Based on comprehensive analyses, we developed an immune subtyping model that classifies extranodal NK/T-cell lymphoma patients into four tumor immune microenvironment subgroups using three immunohistochemical markers (FoxP3, PD-L1, and CD68). The four tumor immune microenvironment subgroups were named immune tolerance, immune evasion-A, immune evasion-B, and immune silenced. The immune tolerance group was characterized by high-Treg counts and was frequently observed in early stage, and nasal extranodal NK/T-cell lymphoma. The immune evasion group showed high cytotoxic T-cell counts and high PD-L1 expression but low Treg counts. In the immune-silenced group, almost all immune responses were exhausted, most patients were at an advanced stage, and had the poorest disease prognosis among the tumor immune microenvironment subgroups. In some patients (n = 3), a shift in the tumor immune microenvironment subgroup classification was observed in sequential biopsies. The response rate to pembrolizumab, an anti-PD-1 antibody, was 100% (1/1) in the immune tolerance group, 60% (3/5) in the immune evasion group, and 0% (0/5) in the immune-silenced group. We classified extranodal NK/T-cell lymphoma into four tumor immune microenvironment subgroups using a new classification system. In conclusion, we propose that the tumor immune microenvironment of extranodal NK/T-cell lymphoma may change during disease progression and may serve as a useful biomarker for immunotherapy.