Description: 〈jats:p〉Abstract. Radiocarbon is a tracer that provides unique insights into the ocean's ability to sequester CO2 from the atmosphere. While spatial patterns of radiocarbon in the ocean interior can indicate the vectors and timescales for carbon transport through the ocean, estimates of the global average ocean–atmosphere radiocarbon age offset (B-Atm) place constraints on the closure of the global carbon cycle. Here, we apply a Bayesian interpolation method to compiled B-Atm data to generate global interpolated fields and mean ocean B-Atm estimates for a suite of time slices across the last deglaciation. The compiled data and interpolations confirm a stepwise and spatially heterogeneous “rejuvenation” of the ocean, suggesting that carbon was released to the atmosphere through two swings of a “ventilation seesaw” operating between the North Atlantic and both the Southern Ocean and the North Pacific. Sensitivity tests using the Bern3D model of intermediate complexity demonstrate that a portion of the reconstructed deglacial B-Atm changes may reflect “phase-attenuation” biases that are unrelated to ocean ventilation and that arise from independent atmospheric radiocarbon dynamics instead. A deglacial minimum in B-Atm offsets during the Bølling–Allerød could partly reflect such a bias. However, the sensitivity tests further demonstrate that when correcting for such biases, ocean “ventilation” could still account for at least one-third of deglacial atmospheric CO2 rise. This contribution to CO2 rise appears to have continued through the Younger Dryas, though much of the impact was likely achieved by the end of the Bølling–Allerød, indicating a key role for marine carbon cycle adjustment early in the deglacial process. Our global average B-Atm estimates place further new constraints on the long-standing mystery of global radiocarbon budget closure across the last deglaciation and suggest that glacial radiocarbon production levels are likely underestimated on average by existing reconstructions. 〈/jats:p〉