Context. Although few in number, high-mass stars play a major role in the interstellar energy budget and the shaping of the Galactic environment; however, the formation of high-mass stars is not well understood, because of their large distances, short time scales, and heavy extinction. Aims. The chemical composition of the massive cores forming high-mass stars can put some constraints on the time scale of the massive star formation: sulfur chemistry is of specific interest thanks to its rapid evolution in warm gas and because the abundance of sulfur-bearing species increases significantly with the temperature. Methods. Two mid-infrared quiet and two brighter massive cores were observed in various transitions (E_(up) to 289 K) of CS, OCS, H_2S, SO, and SO_2 and of their ^(34)S isotopologues at mm wavelengths with the IRAM 30m and CSO telescopes. The 1D modeling of the dust continuum is used to derive the density and temperature laws, which were then applied in the RATRAN code to modeling the observed line emission and to deriving the relative abundances of the molecules. Results. All lines are detected, except the highest energy SO_2 transition. Infall (up to 2.9 km s^(-1)) may be detected towards the core W43MM1. The inferred mass rate is 5.8–9.4 10^(-2) M_⊙/yr. We propose an evolutionary sequence of our sources (W43MM1 → IRAS18264-1152 → IRAS05358+3543 → IRAS18162-2048), based on the SED analysis. The analysis of the variations in abundance ratios from source to source reveals that the SO and SO_2 relative abundances increase with time, while CS and OCS decrease. Conclusions. Molecular ratios, such as [OCS/H_2S], [CS/H_2S], [SO/OCS], [SO_2/OCS], [CS/SO], and [SO_2/SO] may be good indicators of evolution, depending on layers probed by the observed molecular transitions. Observations of molecular emission from warmer layers, so that involving higher upper energy levels must be included.