Evolved Wolf-Rayet stars form a key aspect of massive star evolution, and their strong outflows determine their final fates. In this study, we calculate grids of stellar models for a wide range of initial masses at five metallicities (ranging from solar down to just 2 per cent solar). We compare a recent hydrodynamically-consistent wind prescription with two earlier frequently-used wind recipes in stellar evolution and population synthesis modelling, and we present the ranges of maximum final masses at core He-exhaustion for each wind prescription and metallicity Z. Our model grids reveal qualitative differences in mass-loss behaviour of the wind prescriptions in terms of “convergence”. Using the prescription from Nugis & Lamers the maximum stellar black hole is found to converge to a value of 20-30 $\rm M_{\odot }$, independent of host metallicity, however when utilising the new physically-motivated prescription from Sander & Vink there is no convergence to a maximum black hole mass value. The final mass is simply larger for larger initial He-star mass, which implies that the upper black hole limit for He-stars below the pair-instability gap is set by prior evolution with mass loss, or the pair instability itself. Quantitatively, we find the critical Z for pair-instability (ZPI) to be as high as 50 per cent $\rm Z_{\odot }$, corresponding to the host metallicity of the LMC. Moreover, while the Nugis & Lamers prescription would not predict any black holes above the approx 130 $\rm M_{\odot }$ pair-instability limit, with Sander & Vink winds included, we demonstrate a potential channel for very massive helium stars to form such massive black holes at ∼ 2 per cent $\rm Z_{\odot }$ or below.