Individual sinking slabs present markedly different dip angles between 410 km and 660 km depths, from vertical slabs penetrating the lower mantle to slabs stagnating above the lower mantle. Proposed factors determining these contrasted deep slab dip angles include the magnitude and evolution of trench retreat, mantle viscosity, slab rheology and phase changes. Here we assess the success of paleo-geographically driven global mantle flow models in matching slabs in tomographic models down to 1,000 km depth. We quantify the spatial match between predicted present-day mantle temperature anomalies and vote maps of tomographic models. We investigate the sensitivity of the spatial match to input parameters of the mantle flow model: imposed tectonic reconstruction, model start age, Rayleigh number, viscosity contrast between the upper and lower mantle, and phase changes. We evaluate the visual match between model slabs and tomographic vote maps for three circum-Pacific regions with contrasted deep slab dip angles. The match between predicted model slabs and slabs inferred from tomography can be used to calibrate the Rayleigh number and viscosity contrast between the upper and lower mantle appropriate for our models. The temporal evolution of the models and the global match at present-day suggest that the subduction history could be refined in the global tectonic reconstructions that we considered. For example, subduction to the east of Japan should be offset by approximately 100 km to the west at ~ 80 Ma to match the anchoring of the plate into the lower mantle suggested by tomography.