ABSTRACT The ages of stars have historically relied on isochrone fitting of standardized grids of models. While these stellar models have provided key constraints on observational samples of massive stars, they inherit many systematic uncertainties, mainly in the internal mixing mechanisms applied throughout the grid, fundamentally undermining the isochrone method. In this work, we utilize the mass–lumiosity (M–L) plane of Higgins & Vink as a method of determining stellar age, with mixing-corrected models applying a calibrated core overshooting αov and rotation rate to fit the observational data. We provide multiple test-beds to showcase our new method, while also providing comparisons to the commonly used isochrone method, highlighting the dominant systematic errors. We reproduce the evolution of individual O stars, and analyse the wider sample of O and B supergiants from the VLT-FLAMES Tarantula Survey, providing dedicated models with estimates for αov, Ω/Ωcrit, and ultimately stellar ages. The M–L plane highlights a large discrepancy in the spectroscopic masses of the O supergiant sample. Furthermore the M–L plane also demonstrates that the evolutionary masses of the B supergiant sample are inappropriate. Finally, we utilize detached eclipsing binaries, VFTS 642 and VFTS 500, and present their ages resulting from their precise dynamical masses, offering an opportunity to constrain their interior mixing. For the near-TAMS system, VFTS 500, we find that both components require a large amount of core overshooting (αov ≃ 0.5), implying an extended main-sequence width. We hence infer that the vast majority of B supergiants are still burning hydrogen in their cores.