Major, trace and volatile elements were measured in a suite of primitive macrocrysts and melt inclusions from the thickest layer of the 10 ka Grímsvötn tephra series (i.e. Saksunarvatn ash) at Lake Hvítárvatn in central Iceland. In the absence of primitive tholeiitic eruptions (MgO > 7 wt%) within the Eastern Volcanic Zone (EVZ) of Iceland, these crystal and inclusion compositions provide an important insight into magmatic processes in this volcanically productive region. Matrix glass compositions show strong similarities with glass compositions from the AD 1783–1784 Laki eruption, confirming the affinity of the tephra series with the Grímsvötn volcanic system. Macrocrysts can be divided into a primitive assemblage of zoned macrocryst cores (An78–An92, Mg#cpx = 82–87, Fo79.5–Fo87) and an evolved assemblage consisting of unzoned macrocrysts and the rims of zoned macrocrysts (An60–An68, Mg#cpx = 71–78, Fo70–Fo76). Although the evolved assemblage is close to being in equilibrium with the matrix glass, trace element disequilibrium between primitive and evolved assemblages indicates that they were derived from different distributions of mantle melt compositions. Juxtaposition of disequilibrium assemblages probably occurred during disaggregation of incompatible trace element-depleted mushes (mean La/Ybmelt = 2.1) into aphyric and incompatible trace element-enriched liquids (La/Ybmelt = 3.6) shortly before the growth of the evolved macrocryst assemblage. Post-entrapment modification of plagioclase-hosted melt inclusions has been minimal and high-Mg# inclusions record differentiation and mixing of compositionally variable mantle melts that are amongst the most primitive liquids known from the EVZ. Coupled high-field strength element (HFSE) depletion and incompatible trace element enrichment in a subset of primitive plagioclase-hosted melt inclusions can be accounted for by inclusion formation following plagioclase dissolution driven by interaction with plagioclase-undersaturated melts. Thermobarometric calculations indicate that final crystal–melt equilibration within the evolved assemblage occurred at ~1140 °C and 0.0–1.5 kbar. Considering the large volume of the erupted tephra and textural evidence for rapid crystallisation of the evolved assemblage, 0.0–1.5 kbar is considered unlikely to represent a pressure of long-term magma accumulation and storage. Multiple thermometers indicate that the primitive assemblage crystallised at high temperatures of 1240–1300 °C. Different barometers, however, return markedly different crystallisation depth estimates. Raw clinopyroxene–melt pressures of 5.5–7.5 kbar conflict with apparent melt inclusion entrapment pressures of 1.4 kbar. After applying a correction derived from published experimental data, clinopyroxene–melt equilibria return mid-crustal pressures of 4 ± 1.5 kbar, which are consistent with pressures estimated from the major element content of primitive melt inclusions. Long-term storage of primitive magmas in the mid-crust implies that low CO2 concentrations measured in primitive plagioclase-hosted inclusions (262–800 ppm) result from post-entrapment CO2 loss during transport through the shallow crust. In order to reconstruct basaltic plumbing system geometries from petrological data with greater confidence, mineral–melt equilibrium models require refinement at pressures of magma storage in Iceland. Further basalt phase equilibria experiments are thus needed within the crucial 1–7 kbar range.