The band edges and band gaps of ${(\mathrm{In}\mathrm{As})}_{n}∕{(\mathrm{Ga}\mathrm{Sb})}_{m}$ $(n,m=1,20)$ superlattices have been theoretically studied through the plane-wave empirical pseudopotential method for different situations: (i) different substrates, GaSb and InAs; (ii) different point group symmetries, ${C}_{2v}$ and ${D}_{2d}$; and (iii) different growth directions, (001) and (110). We find that (a) the band gaps for the (001) ${C}_{2v}$ superlattices on a GaSb substrate exhibit a nonmonotonic behavior as a function of the GaSb barrier thickness when the number of ${(\mathrm{In}\mathrm{As})}_{n}$ layers exceed $n=5$; (b) substrate effects: compared with the GaSb substrate, the different strain field generated by the InAs substrate leads to a larger variation of the band gaps for the (001) ${C}_{2v}$ superlattices as a function of the InAs well thickness; (c) effect of the type of interfacial bonds: the In-Sb bonds at the interfaces of the (001) ${D}_{2d}$ superlattices partially pin the band edge states, reducing the influence of the confinement effects on electrons and holes, and lowering the band gaps as compared to the (001) ${C}_{2v}$ case. The valence band maximum of the (001) ${D}_{2d}$ superlattices with Ga-As bonds at the interfaces are shifted down, increasing the band gaps as compared to the (001) ${C}_{2v}$ case; (d) effect of layer orientation: the presence of In-Sb bonds at both interfaces of the (110) superlattices pin the band edge states and reduces the band gaps, as compared to the (001) ${C}_{2v}$ case. An anticrossing between the electron and hole levels in the (110) superlattices, for thin GaSb and thick InAs layers, leads to an increase of the band gaps, as a function of the InAs thickness; (e) superlattices vs random alloys: the comparison between the band edges and band gaps of the superlattices on a GaSb substrate and those for random alloys, lattice matched to a GaSb substrate, as a function of the In composition, shows that the random alloys present almost always higher band gaps and give a clear indication of the effect of superlattice's ordering and period on the behavior of the band gaps and band edges. Inclusion of interfacial interdiffusion, using the approach of Magri and Zunger [Phys. Rev. B 65, 165302 (2002)], is shown to significantly increase the band gaps relative to the predictions for abrupt superlattices, bringing the results closer to experiment. It is noteworthy that $\mathbf{k}\ensuremath{\cdot}\mathbf{p}$ model fit instead measured gaps corresponding to interdiffused interfaces using a chemically abrupt model.