Eocene granite in the Northern Himalayan Gneiss Domes (NHGD), Southern Tibet, offers insights into the early India-Asia collision, yet the source anatectic pressure remains poorly constrained. This study employs zircon U–Pb geochronology, whole-rock geochemistry, and Sr–Nd isotopes on a Late-Eocene granites-intermediate rocks suite from the Yardoi gneiss dome, combined with phase equilibrium calculations on metabasites to unravel their petrogenesis and implications. The results indicate that these rocks crystallized at ca. 43–40 Ma. The granites exhibit high Sr/Y ratios (44–111), are peraluminous and Na-rich. The intermediate rocks have higher Al2O3, CaO, and Na2O than the granites, while other elements remain comparable. Sr–Nd isotope indicates that metabasite is the primary source for both rock types. Phase equilibrium calculations imply that the granites originated from high-pressure melting of amphibolite at 750–800°C and 1.3–1.8 GPa, yielding a plagioclase-depleted residue. Variations in muscovite content in the residua may explain the differing Rb–Ba concentrations observed in the granites. Limited HREE fractionation in the granites may stem from slow element diffusion in subsolidus garnet and low peritectic garnet abundance. Moreover, the intermediate rocks likely formed under similar P–T conditions but from alkaline metasomatized metabasite. We conclude that the decrease of plagioclase, while muscovite remains stable is a common feature during metabasite high-pressure melting. The Eocene primary high Sr/Y magmatism resulted from melting of the thickened lower crust, with high-density residues potentially accumulating and detaching due to gravitational instability, triggering extensional tectonics in the NHGD since ca. 35 Ma.