Diamonds containing fluid inclusions provide invaluable samples of upper mantle fluids, the study of which illuminates not only diamond formation but also the long-term evolution of the subcratonic, lithospheric mantle. The very large range of inclusion compositions worldwide has been interpreted to represent four end-member fluids: saline (rich in Na + K + Cl); silicic (rich in Si + Al); and carbonatitic (rich in Ca + Mg + Fe, with low-Mg and high-Mg end members). However, the sources and evolution of these fluids and the processes involved in diamond formation are still unclear.We used an unusual study of diamonds from the Panda kimberlite (Ekati Mine, Northwest Territories, Canada) in which both mineral and fluid inclusions in the diamonds were analyzed (Tomlinson et al., 2006) to develop models of the saline, silicic, and low-Mg carbonatitic fluids present in the Panda fluid inclusions. The models used aqueous speciation and solubility calculations to link the solid and fluid inclusion chemistry with model upper mantle rock types. We used the extended Deep Earth Water model to calculate equilibrium constants previously calibrated with experimental rock solubilities referring to upper mantle temperatures and pressures (Huang and Sverjensky, 2019). Our results at 950 °C and 4.5 GPa suggest that the saline fluid could originate from peridotite, the silicic fluid from eclogite, and the low-Mg carbonatitic fluid from carbonated dunite.The fluid models were then used to predict the irreversible, chemical mass transfer when the carbonatitic fluid infiltrated a harzburgite containing a saline fluid. Simultaneous reduction of formate and bicarbonate in the carbonatitic fluid and oxidation of aqueous hydrocarbons from the peridotitic fluid during mixing and reaction with harzburgite resulted in the formation of diamond, olivine, garnet, and clinopyroxene, and increases in the logfO2 and pH. Olivine was predicted to become more Fe-rich and garnet more Ca and Fe-rich with reaction progress, in agreement with reported temporal trends (core-to-rim) in the Panda mineral inclusions. The fluid at the site of diamond formation became more saline with reaction progress and the predicted aqueous phase concentrations of all elements changed consistent with trends in Panda fluid inclusions. In contrast, a prediction for a saline fluid infiltrating a harzburgite containing a carbonatitic fluid resulted in trends of the silicate minerals and the salinity with reaction progress that were in the opposite direction to data from the Panda diamonds. Overall, our study strongly supports the notion that fluids from subducting slabs could mix and precipitate diamonds containing carbon from both oxidized and reduced sources, while adding Ca and Fe to the sub-lithospheric cratonic mantle through metasomatic reactions.