This paper discusses the computation and re-computations of the preliminary Philippine Geoid Model 2014 (PGM2014) into PGM2016 and then PGM2018 with the technical assistance of the Technical University of Denmark (DTU-Space) using data from land gravity, airborne gravity, marine satellite altimetry and satellite gravity data from the ESA GOCE mission. Digital terrain models used in the computation process were based on 15” SRTM data, and all data combined with remove-restore state-of-the art geoid determination methods. The preliminary 2014 geoid model was computed in a global vertical reference system with an estimated standard deviation around 0.25 m, then fitted to GNSS/Leveling with an a priori RMS value of only 0.50 m at that point in time, highlighting the challenges in the geodetic infrastructure of the Philippines, with extreme gravity field variations, and geodynamics which can make GNSS-levelling comparison geoid values highly error-prone. To improve the PGM2014, an interim PGM2016 hybrid geoid was computed using reprocessed, re-analyzed and densified land gravity and improved GNSS/Leveling data. Significant improvement was seen in the reprocessed gravity data as well as the final geoid (SD=0.022 m; RMS = 0.054 m) after a more tight GNSS fit, taking into account the different vertical datum used in the islands. In 2018, with new satellite data available, densified gravity data, and additional GNSS/Leveling points were used in a new geoid computation. The new computed geoid has an improved fit to GNSS/Leveling of 2 cm. The formal error estimate of the new hybrid geoid across the Philippines is around 10 cm rms, still limited to some degree by island height system and geodynamics effects.