Laser powder bed fusion (LPBF) of Ti6Al4V alloy is commonly simulated using finite element (FE) models to predict thermal history and residual stress development. However, many numerical studies employ geometries that are much smaller than practical components, raising concerns about the reliability of stress predictions at reduced scales. In this study, a three-dimensional coupled thermo-mechanical FE model is developed in Abaqus to investigate the effect of geometric scaling on residual stress predictions in LPBF-processed Ti6Al4V material. It aims to determine the extent to which the LPBF build geometry can be scaled down while still yielding reliable numerical results at an acceptable computational cost. The objective is to establish a balance between simulation accuracy and computational efficiency through systematic geometric scaling. The model incorporates layer-by-layer element activation as well as temperature-dependent material properties, and a volumetric Gaussian heat source implemented via a DFLUX subroutine. Thermal validation against published data demonstrates good agreement. A systematic scaling study is performed by varying the dimensions of square prints while maintaining identical process parameters and build height. Residual stress distributions along central longitudinal and transverse paths are evaluated using normalized coordinates to enable direct comparison across scales. The results reveal a pronounced size effect, with residual stress profiles converging for print side lengths of 1.0 mm and larger. Below this threshold, stress distributions become size-dependent and potentially misleading. Increasing the model size beyond this threshold leads to a significant rise in computational cost without notable improvement in prediction accuracy, highlighting the trade-off between numerical accuracy and efficiency.