Due to the rapid increasing use of GPS heighting, which already gives the same accuracy as levelling over some 10 to 100 km, there is an urgent need to provide the "cm quasi-geoid" to geodesists and surveyors. In this context, the Geodetic Laboratory of National Centre of Space Techniques has recently focused a part of the current research on the precise geoid determination using different methods. In 1997, the first determination of a preliminary geoid in a small zone in the north of Algeria was calculated by using the least squares collocation and the Gravsoft software. Nowadays, an improved quasi-geoid has been computed over the whole of Algeria. This solution was based on the validated gravity data supplied by the Geophysical Exploration Technology Ltd (GETECH), topographic information and optimal geopotential model, which were combined by using the remove-restore technique in connection with the Fast Fourier Transformation. The GLOBE 30" digital elevation model has provisionally been chosen as the DEM to be used for computation of the effects of the topography according to the RTM reduction modelling method. However, the spherical harmonic coefficient set OSU91A, complete to a degree and an order 360, was adopted as a reference in order to eliminate the long wavelengths of the gravity field. In this paper, the main features of the Algerian quasi-geoid solution are summarized, and extended tests of this solution are undertaken using the new GPS and levelling data collected from the TYRGEONET project and the local GPS/Levelling surveys. The comparisons based on different GPS campaigns provide, after fitting by using the four-parameter transformation, an RMS differences ± 11cm especially for the north part of the country over distances of 1 to 1000 km