Micro-analytical determination of Fe3+/∑Fe ratios in mid-ocean ridge basalt (MORB) glasses using micro X-ray absorption near edge structure (μ-XANES) spectroscopy reveals a substantially more oxidized upper mantle than determined by previous studies. Here, we show that global MORBs yield average Fe3+/∑Fe ratios of 0.16 ± 0.01 (n = 103), which trace back to primary MORB melts equilibrated at the conditions of the quartz–fayalite–magnetite (QFM) buffer. Our results necessitate an upward revision of the Fe3+/∑Fe ratios of MORBs, mantle oxygen fugacity, and the ferric iron content of the mantle relative to previous wet chemical determinations. We show that only 0.01 (absolute, or < 10%) of the difference between Fe3+/∑Fe ratios determined by micro-colorimety and XANES can be attributed to the Mössbauer-based XANES calibration. The difference must instead derive from a bias between micro-colorimetry performed on experimental vs. natural basalts. Co-variations of Fe3+/∑Fe ratios in global MORB with indices of low-pressure fractional crystallization are consistent with Fe3+ behaving incompatibly in shallow MORB magma chambers. MORB Fe3+/∑Fe ratios do not, however, vary with indices of the extent of mantle melting (e.g., Na2O(8)) or water concentration. We offer two hypotheses to explain these observations: The bulk partition coefficient of Fe3+ may be higher during peridotite melting than previously thought, and may vary with temperature, or redox exchange between sulfide and sulfate species could buffer mantle melting at ~ QFM. Both explanations, in combination with the measured MORB Fe3+/∑Fe ratios, point to a fertile MORB source with greater than 0.3 wt.% Fe2O3.