After the GNSS/levelling points validation and analysis, only 341 points were selected for q-geoid model from the total amount of 414 points. Some points were excluded at once after GNSS data processing, because of high RMS error (more than 1.2 cm), some other points were excluded during the iterations made in DFHRS software. The points which exceeded 3 cm reproductivity error were excluded. (Reproductivity means the evaluation of the point in computations if this points would not be taken into account during the determination process of q-geoid). This error shows the real shift between q-geoid and h-H difference. In the result some versions of q-geoid are computed using different types of data, the analysis is provided in publications (SCOPUS, WoS), and the best quasi-model is chosen, which can be further used in different branches: land surveying, road and bridge construction, civil engineering and many other fields, where precise normal heights are needed. GNSS observations can be used for normal height determination by applying GGI q-geoid model, instead of levelling works, as it is time consuming and expensive process.