Accurate organ and cancer segmentation in medical imaging, especially in 3D CT scans, is essential for precise diagnosis, treatment planning, and disease monitoring. The MICCAI FLARE24 challenge aims to advance pan-cancer segmentation algorithms, with Task 1 focusing on whole-body cancer segmentation in CT scans. In this paper, we present an efficient approach utilizing a lightweight 3D U-Net architecture to address this challenge. Our model comprises only four resolution stages and approximately 5.6 million parameters, significantly reducing computational demands while maintaining performance. Our method processed the 279 CT images in the public validation set in just 18 min, averaging under 4 s per scan. For individual cases, the average prediction time per scan was approximately 20 s. We achieved an average Dice Similarity Coefficient (DSC) of 25.34% and a Normalized Surface Dice (NSD) of 24.40% on the public validation set.