Current genomic models in diffuse large B-cell lymphoma (DLBCL) are based on single tumor biopsies, which might underestimate heterogeneity. Data on mutational evolution largely remains unknown. An exploratory study using whole exome sequencing on paired (primary and relapse) formalin fixed paraffin embedded DLBCL biopsies (n = 14) of 6 patients was performed to globally assess the mutational evolution and to identify gene mutations specific for relapse samples from patients treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone. A minority of the mutations detected in the primary sample (median 7.6%, range 4.8–66.2%) could not be detected in the matching relapse sample. Relapsed DLBCL samples showed a mild increase of mutations (median 12.5%, range 9.4–87.6%) as compared to primary tumor biopsies. We identified 264 genes possibly related to therapy resistance, including tyrosine kinases (n = 18), (transmembrane) glycoproteins (n = 73), and genes involved in the JAK-STAT pathway (n = 7). Among the potentially resistance related genes were PIM1, SOCS1, and MYC, which have been reported to convey a risk for treatment failure. In conclusion, we show modest temporal heterogeneity between paired tumor samples with the acquisition of new mutations and identification of genes possibly related to therapy resistance. The mutational evolution could have implications for treatment decisions and development of novel targeted drugs.