Nuclear weapons testing generates large volumes of glassy material that influence the transport of dispersed actinides in the environment and may carry information on the composition of the detonated device. We determine the oxidation state of U and Fe (which is known to buffer the oxidation state of actinide elements and to affect the redox state of groundwater) in samples of melt glass collected from three U.S. nuclear weapons tests. For selected samples, we also determine the coordination geometry of U and Fe, and we report the oxidation state of Pu from one melt glass sample. We find significant variations among the melt glass samples, and in particular, find a clear deviation in one sample from the expected buffering effect of Fe(II)/Fe(III) on the oxidation state of uranium. In the first direct measurement of Pu oxidation state in a nuclear test melt glass, we obtain a result consistent with existing literature that proposes Pu is primarily present as Pu(IV) in post-detonation material. In addition, our measurements imply that highly mobile U(VI) may be produced in significant quantities when melt glass is quenched rapidly following a nuclear detonation, though these products may remain immobile in the vitrified matrices. The observed differences in chemical state among the three samples show that redox conditions can vary dramatically across different nuclear test conditions. The local soil composition, associated device materials, and the rate of quenching are all likely to affect the final redox state of the glass. The resulting variations in glass chemistry are significant for understanding and interpreting debris chemistry, and the later environmental mobility of dispersed material.