Context. The majority of massive stars are part of binary systems. In about a quarter of these, the companions are so close that mass transfer occurs while they undergo core hydrogen burning, first on the thermal and then on the nuclear timescale. The nuclear timescale mass transfer leads to observational counterparts: the semi-detached so-called massive Algol binaries. These systems may provide urgently needed tests of the physics of mass transfer. However, comprehensive model predictions for these systems are sparse. Aims. We use a large grid of detailed evolutionary models of short-period massive binaries, and follow-up population synthesis calculations, to derive probability distributions of the observable properties of massive Algols and their descendants. Methods. Our results are based on ∼10,000 binary model sequences calculated with the stellar evolution code MESA, using a metallicity suitable for the Large Magellanic Cloud (LMC), covering initial donor masses between 10 M and 40 M and initial orbital periods above 1.4 d. These models include internal differential rotation and magnetic angular momentum transport, non-conservative mass and angular momentum transfer between the binary components, and time-dependent tidal coupling. Results. Our models imply ∼30, or ∼3% of the ∼1,000 core hydrogen burning O-star binaries in the LMC to be currently in the semidetached phase. Our donor models are up to 25-times more luminous than single stars of identical mass and effective temperature, which agrees with the observed Algols. A comparison of our models with the observed orbital periods and mass ratios implies rather conservative mass transfer in some systems, while very inefficient one in others. This is generally well reproduced by our spindependent mass transfer algorithm, except for the lowest considered masses. The observations reflect the slow increase of the surface nitrogen enrichment of the donors during the semi-detached phase all the way to CNO-equilibrium. We also investigate the properties of our models after core hydrogen depletion of the donor star, when these models correspond to Wolf-Rayet/helium+OB star binaries. Conclusions. A dedicated spectroscopic survey of massive Algol systems may allow to derive the dependence of the efficiency of thermal timescale mass transfer on the binary parameters, as well as the efficiency of semiconvective mixing in the stellar interior. This would be a crucial step towards reliable binary models up to the formation of supernovae and compact objects.