Residual terrain modeling (RTM) is a technique for determining the high-frequency components of the gravity field, and is widely used within the remove-compute-restore (RCR) scheme for local quasigeoid modeling. RTM consists of computing the gravitational effects generated by the masses located between the topography, described by a digital elevation model (DEM), and a smoother version of it, denoted topographic reference surface. This study aims to compare two RTM approaches in the context of local quasigeoid modeling. The first approach, denoted moving-average (MA), applies a moving-average window to the DEM to obtain the topographic reference surface, and then computes the RTM effects through numerical integration. The second approach, denoted spectral (SP) method, recovers the RTM effects from the ERTM2160 model. Both approaches were applied for quasigeoid modeling with the RCR scheme. Residual height anomalies were determined through Molodensky’s integral and the 1D-FFT method, with a Wong-Gore modification of Stokes’ kernel. Models were determined in five regions surrounding the Argentinean stations of the International Height Reference Frame (IHRF). The XGM2019e global geopotential model (GGM) was used up to degree and order 2159. In each region, the models were compared with one another and validated against GNSS/leveling benchmarks. Results show that the SP approach generates smoother residual gravity anomalies. Differences between quasigeoid models reach a maximum standard deviation of 0.026 m in regions with the highest topography. However, both MA and SP models show a similar performance when compared with GNSS/leveling data. Overall, the SP approach is preferable for local quasigeoid modeling, as it provides smoother residual gravity anomalies and is computationally less demanding.