Although broad consensus exists that photoirradiation of mixed-halide lead perovskites leads to anion segregation, no model today fully rationalizes all aspects of this near ubiquitous phenomenon. Here, we quantitatively compare experimental, CsPb(I 0.5 Br 0.5 ) 3 nanocrystal (NC) terminal anion photosegregation stoichiometries and excitation intensity thresholds to a band gap-based, thermodynamic model of mixed-halide perovskite photosegregation. Mixed-halide NCs offer strict tests of theory given physical sizes, which dictate local photogenerated carrier densities. We observe that mixed-anion perovskite NCs exhibit significant robustness to photosegregation, with photosegregation propensity decreasing with decreasing NC size. Observed size-and excitation intensity-dependent photosegregation data agree with model predicted size-and excitation intensity-dependent terminal halide stoichiometries. Established correspondence between experiment and theory, in turn, suggests that mixed-halide perovskite photostabilities can be predicted a priori using local gradients of (empirical) Vegard's law expressions of composition-dependent band gaps.