Abstract The determination of a high-accuracy and high-resolution geoid is of indispensable importance as it forms one of the cornerstones of a modern geodetic reference system. With that in mind, the State Authority for Geospatial Information (ASIG) of the Republic of Albania has launched an effort to modernize the gravity infrastructure of the country and as a consequence to make available a new gravimetric geoid model of enhanced resolution and accuracy for the country. In that respect, the necessary procedures to establish the primary and secondary leveling network have been detailed, collocated with gravity measurements, as well as the determination of a preliminary gravimetric geoid based on existing and new gravity data. In this work, we first focus on the gravity campaigns, data processing, reductions, and adjustments carried out as part of establishing the reference benchmarks of the second-order gravity network. Then the methodological steps followed for the determination of the geoid on a 1′ × 1′ grid over Albania, in an area bounded between 38.5° ≤ φ ≤ 43.5° and 18.0° ≤ λ ≤ 22.0°, are outlined. For the geoid determination, the remove-compute-restore (RCR) concept is employed using two global geopotential models (GGMs), namely EIGEN6c4 and XGM2019e, and two approaches for the topographic effects through a residual terrain model (RTM) correction with the spectral and classical computation. Then, the estimation of the geoid is carried out using least squares collocation (LSC) and fast Fourier transform (FFT) methods, and the validation is performed against both GNSS/leveling data over Albania and a high-accuracy leveling traverse connecting Durrës and Tirana. From the results achieved, the FFT-based geoid presented a mean value of the differences at the 3 cm level and an std of 14.1 cm. When compared to the new leveling traverse, the new geoid has shown differences at the ∼16 cm level which indicates the bias between the new gravimetric geoid from FFT and the Albanian vertical datum. The relative differences over the measured leveling BMs are at the 1–7 mm level. A first-guess hybrid geoid was also determined, after the parametric fit with the third-order polynomial model, reaching the 12.7 cm agreement with the old vertical datum, which is very satisfactory considering the fact that the acquired land gravity data manage to cover only a small portion of the gravity variability due to the long distances between them.