As the number of Earth geopotential models (EGM) grows with the increasing number of data collected by dedicated satellite gravity missions, CHAMP, GRACE and GOCE, measuring the differences among the models and monitoring the improvements in gravity field recovery are required. This study assesses the performance of recent EGMs derived from CHAMP, GRACE, and other data in comparison to the earlier two models in Turkish territory. Also the improvement capacities of these EGMs using local terrestrial data are inspected with two different approaches. In the first approach, the spherical harmonic coefficients of EGMs are modified depending on the local gravity data. In the second part, the original EGMs with their maximum harmonic expansions are employed in Remove Compute Restore algorithm for high resolution local geoid modelling. The assessment results with the local terrestrial data exhibited large disagreements among the models in Turkey. The outputs from regional improvements of EGMs using the gravity observations clarified the important role of the ground truth data contribution in geopotential models. The validations of the computed high resolution geoids rely on the EGMs at the GNSS/levelling networks, having 451 and 309 benchmarks, provided an absolute accuracy by means of standard deviations of the geoid height differences around 20 cm. These results from the final stage of the case study provide a comparison among the EGMs by means of their fit to the local gravity field.