Diffuse large B-cell lymphoma (DLBCL) presents considerable clinical challenges due to its aggressive nature and diverse clinical progression. New molecular biomarkers are urgently needed for outcome prediction. We analyzed blood samples from DLBCL patients and healthy individuals using short, non-coding RNA sequencing. A classifier based on six tsRNAs was developed through random forest and primary component analysis. This classifier, established using Cox proportional hazards modeling with repeated 10-fold cross-validation on an internal cohort of 100 samples analyzed using RT-qPCR, effectively identified high-risk patients with significantly lower overall survival compared to low-risk patients (Hazard ratio: 6.657, 95%CI 2.827-15.68, P = 0.0006). Validation in an external cohort of 160 samples using RT-qPCR confirmed the classifier's robust performance. High-risk status was strongly associated with disease histological subtype, stage, and International Prognostic Index scores. Integration of the classifier into the IPI model enhanced the precision and consistency of prognostic predictions. A dynamic study revealed that patients experiencing a 1.06-fold decrease after one therapy cycle (early molecular response) exhibited better treatment outcomes and prognosis. Furthermore, the 6-tsRNA signature accurately differentiated healthy individuals from DLBCL (AUC 0.882, 95%CI 0.826-0.939). These findings underscore the potential of the identified 6-tsRNA profile as a biomarker for monitoring treatment effectiveness and predicting DLBCL outcomes.