Since Global Leadership Initiative on Malnutrition (GLIM) method was proposed, few studies have applied these new criteria to hematological tumors. In this study, we explored the prevalence of malnutrition according to the GLIM criteria and scored Patient‐Generated Subjective Global Assessment (sPG‐SGA) and their association with 1‐year, 3‐year and 5‐year mortality among patients with non‐Hodgkin's lymphoma (NHL). Malnutrition of all patients were assessed by GLIM criteria and sPG‐SGA. Relationship between the malnutrition based on GLIM criteria or sPG‐SGA and mortality was investigated by Cox regression analyses. The performance of GLIM criteria was evaluated by assessing the sensitivity, specificity, k‐value, receiver operating characteristic (ROC) curve and time‐dependent ROC. Of 963 patients with NHL, the prevalence of malnutrition was 38.8% with GLIM criteria, 65.3% with GLIM‐omitted NRS‐2002 and 53.2% with sPG‐SGA. In comparison with sPG‐SGA, the sensitivity of GLIM criteria was 61.7%, the specificity was 84.8%, and the agreement was moderate (k = 0.48, p < 0.001). Malnutrition based on GLIM criteria could also predict 3‐year and 5‐year mortality after adjusting for confounders, except for sPG‐SGA (HR = 1.816, 95%CI = 1.274–2.589, p = 0.001 for 3‐year mortality; HR = 1.707, 95%CI = 1.223–2.382, p = 0.002 for 5‐year mortality). For patients with NHL, GLIM criteria could be applied as an effective replacement to sPG‐SGA for nutrition assessment and mortality prediction, especially for predicting long‐term prognostic outcomes.