Currently, several gene‐expression signatures that were used to predict survival of diffuse large B‐cell lymphoma (DLBCL) patients, showed a restriction on the practical work for lack of convenient operation. In this study, we screened inflammatory genes whose expression correlated with survival of DLBCL and established a predictive model including IL6, IL1A and CSF3 through multivariate Cox regression based on the expression of these three genes. We validated the model at protein level in our clinical serum cohort composed of 101 patients of DLBCL and 50 healthy controls and 534 DLBCL patients at mRNA level from three independent Gene Expression Omnibus (GEO) data sets. We found our model to be independent of the International Prognostic Index (IPI), moreover, it can augment the predictive power of IPI. In summary, our three‐gene model is sufficient to predict survival of DLBCL patients via measuring the concentration of three inflammatory cytokines in peripheral blood.