Abstract <p>The solubility of platinum and palladium in two phases of low-water fluid (NaCl melt and CO₂–CO–Cl₂(H₂O)) was experimentally studied at <i>P–T</i> parameters of the cumulus of ultramafic–mafic intrusions and low oxygen fugacity, <i>f</i>O₂ = CCO. The solubility of platinum as chloride in NaCl melt with several percent of dissolved carbon dioxide at <i>T</i> = 950°C and <i>P</i> = 200 MPa and increased chlorine fugacity maintained by the Ag–AgCl buffer is 23 ± 4 ppm, and that of palladium is 74 ± 26 ppm. At the same <i>P–T</i> parameters and chlorine fugacity at the FeCl₂–FeCl₃ buffer, the solubility of native platinum in CO₂–CO–Cl₂ fluid in the form of carbonyl and possibly mixed chloride–carbonyl complexes is 341 ± 110 ppm, and that of palladium was determined less accurately as 335 ppm. At high oxygen fugacity, the solubility of palladium in silicate melt, hydrothermal solution, and supercritical fluid is several orders of magnitude (2 to 4.5) higher than that of platinum. The solubility of native palladium in reduced fluid similar to or slightly lower than that of platinum may explain the elevated Pt/Pd ratios in some PGE ores, despite the similar mean contents of these elements in the primitive mantle and parent magmas of ultramafic–mafic intrusions. Estimates of the reduced fluid–sulfide melt partition coefficients based on the obtained solubility values give at the maximum CO content <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({\text{Kd}}_{{fl - slf}}^{{{\text{Pt}}}}\)</EquationSource> <!--PetrEng2670040Simakin-m1--> </InlineEquation> = 0.47–6.5 and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({\text{Kd}}_{{fl - slf}}^{{{\text{Pd}}}}\)</EquationSource> <!--PetrEng2670040Simakin-m2--> </InlineEquation> = 0.02–0.13, which suggests a possible mechanism for the separation of platinum and palladium in the early postmagmatic process in sulfide-bearing cumulus with the participation of reduced carbon dioxide fluid at high chlorine fugacity.</p>