Recent advances in high-precision potassium (K) isotopic analysis have found considerable isotopic variation in rock samples of the Earth’s continental and oceanic crusts; however, it is still uncertain whether there is any resolvable inter-mineral and mineral-melt K isotopic fractionation during igneous and metamorphic processes. Here, we report K isotope compositions of mineral separates from three extremely well-preserved igneous rocks (intrusive/extrusive and mafic/intermediate/felsic) in order to investigate possible inter-mineral and mineral-melt K isotopic fractionation at magmatic temperatures. For the first time, we found large inter-mineral fractionation of K isotopes in natural samples (up to 1.072‰), where plagioclase displays a significant enrichment of heavier K isotopes compared to potassium feldspar and biotite in granite. In addition, we also observed smaller but measurable K isotope fractionation (0.280 ± 0.030‰) between ternary feldspar phenocrysts and matrices in a trachyandesite, as well as a comparable isotope fractionation (0.331 ± 0.010‰) between plagioclase and the bulk in a gabbroic intrusive rock. We also evaluated such results by comparing the theoretically calculated equilibrium K isotope fractionation factors between relevant igneous minerals in the literature and this study. In general, the measured inter-mineral fractionations are consistent with the theoretical calculations (i.e., plagioclase is enriched in heavier isotopes compared to potassium feldspar). Specifically, the measured K isotope fractionation between the phenocryst rim and matrix in the trachyandesite agrees well with the calculated equilibrium isotope fractionation. However, the measured K isotope fractionations between the phenocryst core and matrix as well as between plagioclase and K-feldspar are significantly larger (by a factor of ∼2–3) than the calculated isotope fractionations, which suggest isotopic disequilibrium due to kinetic processes. Using a range of plagioclase-melt isotope fractionation factors inferred from the theoretical calculations in this study, we modeled the K isotopic fractionation during the formation of lunar anorthositic crust, and the result shows a negligible effect on the K isotopic compositions in both lunar crust and mantle. The K isotopic difference between the Earth and the Moon, therefore, cannot be the result of lunar magma ocean differentiation. Finally, we evaluate the effect of observed inter-mineral fractionations on K–Ar and 40Ar–39Ar dating. This study indicates that the variation of the 40K/K ratio would contribute a maximum 0.08% error to the K–Ar and 40Ar–39Ar age uncertainties. We propose a refined 40K/total K ratio as 0.00011664 ± 0.00000011 (116.64 ± 0.11 ppm) instead of the conventional value, 0.0001167(2), for the present Earth. Because some minerals fractionate K isotopes, ultrahigh-precision age dating with the K–Ca–Ar dating systems must measure the K isotope fractionation in the same mineral fractions used for age dating.