Crystal zoning as well as temperature and pressure estimates from phenocryst phase equilibria are used to constrain the architecture of the intermediate-sized magmatic system (some tens of km³) of Volcán Quizapu, Chile, and to document the textural and compositional effects of magma mixing. In contrast to most arc magma systems, where multiple episodes of open-system behavior obscure the evidence of major magma chamber events (e.g. melt extraction, magma mixing), the Quizapu magma system shows limited petrographic complexity in two large historical eruptions (1846–1847 and 1932) that have contrasting eruptive styles. Quizapu magmas and peripheral mafic magmas exhibit a simple binary mixing relationship. At the mafic end, basaltic andesite to andesite recharge magmas complement the record from peripheral cones and show the same limited range of compositions. The silicic end-member composition is almost identical in both eruptions of Quizapu. The effusive 1846–1847 eruption records significant mixing between the mafic and silicic end-members, resulting in hybridized andesites and mingled dacites. These two compositionally simple eruptions at Volcán Quizapu present a rare opportunity to isolate particular aspects of magma evolution—formation of homogeneous dacite magma and late-stage magma mixing—from other magma chamber processes. Crystal zoning, trace element compositions, and crystal-size distributions provide evidence for spatial separation of the mafic and silicic magmas. Dacite-derived plagioclase phenocrysts (i.e. An_25–40) show a narrow range in composition and limited zonation, suggesting growth from a compositionally restricted melt. Dacite-derived amphibole phenocrysts show similar restricted compositions and furthermore constrain, together with more mafic amphibole phenocrysts, the architecture of the magmatic system at Volcán Quizapu to be compositionally and thermally zoned, in which an andesitic mush is overlain by a homogeneous dacitic magma that is the source for most of the 1846–1847 and 1932 erupted magmas. Dacite formation is best explained by mineral–melt separation (crystal fractionation) from an andesitic mush, which is inferred to have thermally and compositionally buffered the dacite magma thereby keeping it at relatively low crystallinity (<30 vol. %). The dominant cause of compositional diversity is melt separation. Back-mixing of mush (i.e. crystals with signatures of growth both in the andesitic mush and in the dacite magma) into the overlying dacite magma is rarely observed. Recharge events that increase crystal and magma diversity in the dacite magma are limited to an episode of mafic recharge and mixing just prior to the 1846–1847 eruption, where evidence for magma mixing is present on all scales. Chamber-wide mixing was incomplete (mixing efficiency of ∼0·53–0·85) as flow lobes vary significantly in composition along the proposed mixing array. Estimates of viscosity variations during the course of magma mixing suggest that mixing dynamics and the degree of magma interaction on all scales were established at the beginning of the recharge event.