The main purpose of this paper is to present the results of external’ quality tests for global, regional and local geoid models. In this context, available geoid models were evaluated for the height transformation in three different geographic locations in Turkiye on the basis of ground truth GNSS/Leveling data. The differences between observed and computed quantities were investigated. The comparison results of the global, regional and local geoid models for three test areas were presented. The tested geoids are the national geoid model Turkiye-Geoid2003 (TG-03) released by the General Command of Mapping in 2003 which was computed using land and sea gravity data, topographic heights from digital terrain model and GPS/Leveling data, including EGM96 global geopotential model for long wavelengths, the ultra-high resolution model EGM2008 that was released by the US National Geospatial Intelligence Agency, the refined versions of these two geoid models and the GNSS/Leveling derived geometric models produced by the metropolitan municipalities. The comparison of results show that TG03 and EGM2008 models provide about 2 times lower accuracy than the precise local geoid models. In addition, the RMS values of the other two methods (IMPTG03 and IMPEGM2008) compared in the study were obtained as 2.45cm (Istanbul), 5.09cm (Izmır) and 4.77cm (Bursa) for IMPTG03, 2.61cm (Istanbul), 5.45cm (Izmır) and 5cm (Bursa) for IMPEGM2008. It seems that the results are sufficient for many engineering applications in the local areas and EGM2008 can be reach the accuracy of regional or local geoid models after the improvement procedure.