This work presents a comprehensive analysis of the feasibility of using T1-GD (contrast-enhanced T1) MRI sequences for 3D glioblastoma segmentation. The experiments were conducted through the benchmark nnU-Net architecture, considering both MNI152 and SRI24 spaces. In order to promote reproducibility, the open-access UPenn-GBM dataset is used, comprising 147 patients split into training, validation, and held-out test subsets. A K-fold cross-validation is conducted to report robust and reliable results. Additionally, a private cohort composed by 21 patients were included with the aim to test the generalization capability of the approach proposed with different unseen data. The use of SRI24 shows slight gains, with DCS values of 0.893 and 0.819 on the held-out test subsets from the UPenn-GBM and external datasets, respectively. The promising results obtained support the generalization capability of the proposed strategy with unseen data, independent of the data source, and underscore the potential of using only T1-GD sequences as the initial, fast, and simple pulse for screening patients in real-world scenarios where glioblastoma lesions are manually outlined for radiotherapy. Reducing the number of MRI sequences needed decreases acquisition time and improves patients’ well-being, while also reducing computational cost.