<p>Resolving the debate on whether magmas are saturated solely in chromite or if chromite reaches a cotectic with other liquidus phases before segregating is crucial to understanding the origin of stratiform chromitite. Probing Mg isotopic variations provides insight into this issue, as chromite crystallization significantly fractionates melt δ²⁶Mg. Here, we present δ²⁶Mg data for whole rocks from the Stillwater Complex to assess their Mg isotopic evolution. The Banded Series are isotopically homogeneous with δ²⁶Mg averaging − 0.29 ± 0.05‰ (<i>n</i> = 10), which is lighter than komatiitic melts. Rayleigh distillation modeling shows that generating this signature via chromite-only saturation would require 15 wt% crystallization. In contrast, cotectic crystallization of chromite and silicates rapidly depletes the melt’s Mg budget, necessitating only ~ 1.7 wt% chromite to produce the same isotopic shift. This strongly supports a model where chromitite layers form by mechanical sorting of chromite from a cotectic crystal mush. Olivine cumulates and granular harzburgites have average δ²⁶Mg values of -0.17‰, which are isotopically heavier than poikilitic harzburgites with δ²⁶Mg values of -0.32‰. Chromitite samples generally exhibit lighter Mg isotopes, with a wider range from − 0.32 ± 0.03‰ to + 0.03 ± 0.04‰. Modeling results suggest that fluid migration between chromitite and silicate cumulates does not cause their Mg isotopic heterogeneity but rather serves as a critical medium for mechanical sorting. The extensive occurrence of Mg isotopic heterogeneity in Ultramafic Series implies that effects of fluid migration need to be considered in studies of basaltic magma chamber processes, evolution, and oxide ore formation.</p>