A new gravimetric Co-geoid is produced for the Malaysian peninsula making it one of the most recent contributions to Co-geoid modelling in the region. Several data sources are included for the computations; these include 7447 terrestrial free-air gravity anomalies, 10243 altimeter derived gravity anomalies and 692881 mean digital elevation heights. The EGM96 global geopotential model to degree 40 is used in order to determine the long wavelength effect of the Co-geoid surface. Co-geoid undulations are determined through implementation of the remove-restore technique by combining the above-mentioned data sets with the geopotential model in a specific manner. The computations have been done using a two-dimensional spherical Fast Fourier Transform (FFT) with 100% zero padding. Terrain corrections are also computed using the FFT method and applied in the remove-restore technique. The absolute accuracy estimates are given for the Co-geoid undulations using 95 GPS control stations co-located with third order levelling. The comparisons show that using the Wong and Gore (1969) modification of Stoke’s function for the long wave length effect using EGM96 up to degree/order 360 removed, and the modified Stoke’s function with L = 40 and a spherical cap radius of 1.6 ◦ , with data padding for areas of no gravity data produces a Co-geoid which gives a bias of 0.378 meters and a standard deviation of ±0.277 meters respectively. Where no data padding is used a Co-geoid is produced which gives a bias of 0.439 meters and a standard deviation of ±0.284 meters respectively, thus showing the improvement of the Co-geoid solution through the data padding.