Glass Mountain, California, consists of >50 km3 of high-silica rhyolite lavas and associated pyroclastic deposits that erupted over a period of >1 my preceding explosive eruption of the Bishop Tuff and formation of the Long Valley caldera at 0.73 Ma. These “minimum-melt” rhyolites yield Fe-Ti-oxide temperatures of 695–718°C and contain sparse phenocrysts of plagioclase+quartz+magnetite+apatite±sanidine, biotite, ilmenite, allanite, and zircon. Incompatible trace elements show similar or larger ranges within the Glass Mountain suite than within the Bishop Tuff, despite a much smaller range of major-element concentrations, largely due to variability among the older lavas (erupted between 2.1 and 1.2 Ma). Ratios of the most incompatible elements have larger ranges in the older lavas than in the younger lavas (1.2–0.79 Ma), and concentrations of incompatible elements span wide ranges at nearly constant Ce/Yb, suggesting that the highest concentrations of these elements are not the result of extensive fractional crystallization alone; rather, they are inherited from parental magmas with a larger proportion of crustal partial melt. Evidence for the nature of this crustal component comes from the presence of scarce, tiny xenocrysts derived from granitic and greenschist-grade metamorphic rocks. The wider range of chemical and isotopic compositions in the older lavas, the larger range in phenocryst modes, the eruption of magmas with different compositions at nearly the same time in different parts of the field, and the smaller volume of individual lavas suggest either that more than one magma body was tapped during eruption of the older lavas or that a single chamber tapped by all lavas was small enough that the composition of its upper reaches easily affected by new additions of crustal melts. We interpret the relative chemical, mineralogical, and isotopic homogeneity of the younger Glass Mountain lavas as reflecting eruptions from a large, integrated magma chamber. The small number of cruptions between 1.4 and 1.2 ma may have allowed time for a large magma body to coalesce, and, as the chamber grew, its upper reaches became less affected by new inputs of crustal melts, so that trace-element trends in magmas erupted after 1.2 Ma are largely controlled by fractional crystallization. The extremely low Sr concentrations of Glass Mountain lavas imply extensive crystallization in chambers at least hundreds of cubic kilometers in volume. The close similarity in Sr, Nd, and Pb isotopic ratios between the younger Glass Mountain lavas and unaltered Bishop Tuff indicates that they tapped the same body of magma, which had become isotopically homogenous by 1.2 Ma but continued to differentiate after that time. From 1.2 to 0.79 Ma, volumetric eruptive rates may have exceeded rates of differentiation, as younger Glass Mountain lavas become slightly less evolved with time. Early-erupted Bishop Tuff is more evolved than the youngest of the Glass Mountain lavas and is characterized by slightly different trace element ratios. This suggests that although magma had been present for 0.5 my, the composiional gradient exhibited by the Bishop Tuff had not been a long-term, steady-state condition in the Long Valley magma chamber, but developed at least in part during the 0.06-my hiatus between extrusion of the last Glass Mountain lava and the climactic eruption.