The inclusion of rotation in massive star models improves the agreement between theory and observations on at least three important points: 1) rotational mixing allows to produce variations of the surface abundances already during the Main-Sequence phase as is observed. The changes of the surface abundances are more important when, for a given initial velocity, the initial masses are larger, and/or the metallicities are lower; 2) the observed number of red supergiants at the metallicity of the Small Magellanic Cloud (SMC) can be accounted for; 3) the observed variation of the number ratio of Wolf-Rayet to O-type stars as a function of the metallicity can be reproduced. For all these comparisons non-rotating models give unsatisfactory fits. Rotating models results also give interesting insights on questions such as the origin of Be stars, the mechanisms responsible for the huge mass loss rates undergone by the Luminous Blue Variables, the rotation rates of pulsars, the progenitors of collapsars and the sources of primary nitrogen at low metallicity.