Natural killer (NK)/T-cell lymphoma is a rare neoplasm with heterogeneous clinical behaviors. This study aimed to devise a prognostic model specifically for extranodal NK/T-cell lymphoma, providing risk stratification in affected patients. A total of 146 patients newly diagnosed with extranodal NK/T-cell lymphoma were retrospectively analyzed. Independent predictors of survival were determined by Cox regression analysis. The estimated 5-year overall survival rate for all patients was 41.9%. Both the International Prognostic Index and Korean NK/T-cell lymphoma Prognostic Index were prognostic in univariate analysis. The majority of our patients were in the low-risk International Prognostic Index and Korean NK/T-cell lymphoma Prognostic Index category (with no or one adverse factor), but both of these two prognostic models had no discriminating power within the subgroup of these patients. Absolute lymphocyte count at diagnosis, B symptoms, and advanced stage were independently predicted poorer survival. Cox analysis yielded a novel prognostic model based on these three parameters that categorized patients into one of three risk groups: Group 1 (57 cases, 39.0%), no adverse factors; Group 2 (43 cases, 29.5%), one adverse factor; and Group 3 (46 cases, 31.5%), two or three adverse factors. Five-year overall survival was 71.6% for Group 1, 55.5% for Group 2, and 0% for Group 3 (P < 0.0001). The novel prognostic model balanced the distribution of patients into different risk groups with better predictive discrimination. It may be useful to stratify patients with extranodal NK/T-cell lymphoma into different risk groups and provide more information for treatment selection.