A 3-year experiment was established in which four wheat classes were evaluated including soft, medium hard, hard and durum wheat with the objective to determine which class is more efficient in water use under reduced irrigation. The experiments were established during three growing seasons (2016–2018). The amount of water applied were: 26, 34 and 54 cm distributed in 2, 3 and 5 irrigations, respectively. Eighteen genotypes from each wheat class were evaluated in an alpha lattice design with three replicates. Phenological data, yield and yield components were analyzed. Yield in the two-irrigation regime ranged from 3974 to 5436, 4453 to 6909 under three and 6177 to 9107 kg ha−1 under five. Correlation analysis showed that with two irrigations there is a greater association of grain yield with thousand kernel weight (TKW), but under three and five irrigations, grain yield was associated to a greater degree with kernel number per unit area (KNO). When analyzing the grain yield, it was observed that under reduced irrigation (2 and 3 irrigations), bread wheats were superior to durum wheats regardless of the class. Under five irrigations, durum wheats showed the highest yield (8303 kg ha−1); however, they were only significantly superior (Tukey ≤ 0.05) to the hard wheats (7721 kg ha−1). In general, the tested wheats showed higher water efficiency (considered as water productivity) under reduced irrigation than under normal irrigation. The lowest losses in water productivity when going from two to five irrigations were observed in durum wheats (0.17 kg m−3) and the highest losses in the hard wheats (0.38 kg m−3).