Zn/FeT (FeT=Fe2+ + Fe3+) ratios in primitive melts have been proposed as a redox proxy to assess the redox states of the upper mantle. However, to effectively use the melt Zn/FeT ratio as a redox proxy, it is necessary to compare variations of melt Zn/FeT ratios induced by changes in oxygen fugacity (fO2) with variations due to changes in Zn-Fe contents and mineralogy of the sources. Here we show that the melt Zn/FeT ratio variation caused by fO2 change can be expressed as Δ(Znm/FemT) = (Znper/FeperT)* Δ(Fem2+/FemT)/(DZnper/m/DFe2+per/m). Zn/FeT ratios in most arc and MORB peridotites (Znper/FeperT) are 9.0 ± 1.0*10−4, and melt Fe2+/FeT ratio variation resulted from fO2 change [Δ(Fem2+/FemT)] can be easily obtained by the existing model. Hence, if the Zn and Fe2+ partition coefficients (DZnper/m and DFe2+per/m) between melt and peridotite are known, the melt Zn/FeT ratio variation resulting from fO2 change can be estimated. In this study, we determined DZn between olivine, orthopyroxene, clinopyroxene and basaltic melts at 0.75–2.5 GPa and 1250–1450 ℃. Our data show that melt composition (expressed as MgO content) dominantly controls the mineral-melt Zn partitioning under peridotite melting conditions. These data, along with published mineral-melt DFe2+ data, enable us to calculate appropriate DZnper/m/DFe2+per/m, which is 0.78 ± 0.02 under arc and MORB spinel peridotite melting conditions. Based on these parameters, the calculated average melt Zn/FeT ratio variation caused by per log unity fO2 change in typical fO2 span of arc and MORB mantles is only ∼ 0.69 ± 0.20*10−4. Alternatively, melt Zn/FeT ratio variations caused by changes in Zn-Fe contents and mineralogy of peridotites are ±1.10*10−4. These findings indicate that melt Zn/FeT ratio variation induced by one log units fO2 change is on the same order of magnitude as those caused by changes in Zn-Fe contents and mineralogy of peridotites. Therefore, unless the precise chemical and mineralogical compositions of the mantle sources can be determined independently, the melt Zn/FeT redox proxy is unsuitable for tracing the mantle fO2. Our study, combined with recent works on melt Cu and V/Sc redox proxies, suggests that these redox proxies, previously considered as the strongest evidence for a reduced arc mantle, might not support the idea of a reduced arc mantle.