Iron oxide–apatite (IOA or Kiruna-type) deposits typically consist of a magnetite-apatite-actinolite/diopside assemblage and are spatially associated with extensive Na-(Ca) alteration and brecciation. The origin of these deposits is highly controversial and has been ascribed to the separation of iron-oxide/sulfate-(carbonate) melts, magnetite emulsions, or metasomatic replacement by aqueous fluids from silicate magmas. Here, we propose a new model based on the findings from a cluster of IOA deposits located in the early Cretaceous Ningwu andesitic volcanic field, eastern China. In these deposits, magnetite coeval with apatite and actinolite occurs as coarse-grained veins, massive replacement, and fine-grained disseminations in the albitized, often brecciated, apical zones of diorite porphyry intrusions, the overlying andesites, and adjacent sedimentary rocks. The primary magnetite grains from ores with various textures contain similar and variable trace element compositions with up to 5 wt% Ti + V and show the characteristics of high-temperature hydrothermal magnetite in magmatic-hydrothermal systems. Diopside and garnet as well as magnetite contain fluid inclusions with multiple daughter minerals (vapor + halite + sylvite ± anhydrite ± iron chloride ± liquid ± hematite), which show extremely high salinities of more than ~ 90 wt% NaCl_equiv, homogenization temperatures of 745–846 °C, and Cl/Br mole ratios of 2000–6000. In combination with oxygen isotopes of the magnetite-apatite assemblage and the association with shallow-seated ore-hosting porphyry, available evidence suggests that these deposits formed from hydrosaline liquid exsolved from subvolcanic dioritic magmas with high Cl/H₂O at magmatic temperatures (~ 800 °C). Decompression from lithostatic to hydrostatic condition and the interaction with country rocks explain the abundance of breccia bodies and widespread sodic alteration in IOA deposits.