Background: We evaluated patterns of tumor growth in patients with newly diagnosed MGMT-non-methylated glioblastoma who were assigned to undergo radiotherapy in conjunction with bevacizumab/irinotecan (BEV/IRI) or standard temozolomide (TMZ) within the randomized phase II GLARIUS trial. Methods: In 142 patients (94 BEV/IRI, 48 TMZ), we reviewed magnetic resonance imaging scans at baseline and first tumor recurrence. Based on contrast-enhanced T1-weighted and fluid-attenuated inversion recovery images, we assessed tumor growth patterns and tumor invasiveness. Tumor growth patterns were classified as either multifocal or local at baseline and recurrence; at first recurrence, we additionally assessed whether distant lesions appeared. Invasiveness was determined as either diffuse or non-diffuse. Associations with treatment arms were calculated using Fisher’s exact test. Results: At baseline, 115 of 142 evaluable patients (81%) had a locally confined tumor. Between treatment arms, there was no significant difference in the fraction of tumors that changed from an initially local tumor growth pattern to a multifocal pattern (12 and 13%, p = 0.55). Distant lesions appeared in 17% (BEV/IRI) and 13% (TMZ) of patients (p = 0.69). 15% of patients in the BEV/IRI arm and 8% in the TMZ arm developed a diffuse growth pattern from an initially non-diffuse pattern (p = 0.42). Conclusions: The tumor growth and invasiveness patterns do not differ between BEV/IRI and TMZ-treated MGMT-non-methylated glioblastoma patients in the GLARIUS trial. BEV/IRI was not associated with an increased rate of multifocal, distant, or highly invasive tumors at the time of recurrence.