Abstract Glasses produced from decompression experiments conducted by Fiege et al. (2014a) were used to investigate the fractionation of sulfur isotopes between fluid and andesitic melt upon magma degassing. Starting materials were synthetic glasses with a composition close to a Krakatau dacitic andesite. The glasses contained 4.55–7.95wt% H2O, ∼140 to 2700ppm sulfur (S), and 0–1000ppm chlorine (Cl). The experiments were carried out in internally heated pressure vessels (IHPV) at 1030°C and oxygen fugacities (fO2) ranging from QFM+0.8 log units up to QFM+4.2 log units (QFM: quartz–fayalite–magnetite buffer). The decompression experiments were conducted by releasing pressure (P) continuously from ∼400MPa to final P of 150, 100, 70 and 30MPa. The decompression rate (r) ranged from 0.01 to 0.17MPa/s. The samples were annealed for 0–72h (annealing time, tA) at the final P and quenched rapidly from 1030°C to room temperature (T). The decompression led to the formation of a S-bearing aqueous fluid phase due to the relatively large fluid–melt partitioning coefficients of S. Secondary ion mass spectrometry (SIMS) was used to determine the isotopic composition of the glasses before and after decompression. Mass balance calculations were applied to estimate the gas–melt S isotope fractionation factor αg-m. No detectable effect of r and tA on αg-m was observed. However, SIMS data revealed a remarkable increase of αg-m from ∼0.9985±0.0007 at >QFM+3 to ∼1.0042±0.0042 at ∼QFM+1. Noteworthy, the isotopic fractionation at reducing conditions was about an order of magnitude larger than predicted by previous works. Based on our experimental results and on previous findings for S speciation in fluid and silicate melt a new model predicting the effect of fO2 on αg-m (or Δ34Sg–m) in andesitic systems at 1030°C is proposed. Our experimental results as well as our modeling are of high importance for the interpretation of S isotope signatures in natural samples (e.g., melt inclusions or volcanic gases).