Abstract. Despite intense focus on the ~ 190 permil drop in atmospheric Δ14C across the deglacial “mystery interval”, the specific mechanisms responsible for the apparent Δ14C excess in the glacial atmosphere have received considerably less attention. The computationally efficient Bern3D earth system model of intermediate complexity, designed for long-term climate simulations, allows us to address a very fundamental but still elusive question concerning the atmospheric Δ14C record: How can we explain the persistence of relatively high Δ14C values during the millennia after the Laschamp event? Large uncertainties in the pre-Holocene 14C production rate, as well as in the older portion of the Δ14C record, complicate our qualitative and quantitative interpretation of the glacial Δ14C elevation. Here we begin with sensitivity experiments that investigate the controls on atmospheric Δ14C in more idealized settings. We show that the long-term process of sedimentation may be much more important to the simulation of Δ14C than had been previously thought. In order to provide a bounded estimate of glacial Δ14C change, the Bern3D model was integrated with five available estimates of the 14C production rate as well as reconstructed and hypothesized paleoclimate forcing. Model results demonstrate that none of the available reconstructions of past changes in 14C production can reproduce the elevated Δ14C levels during the last glacial. In order to increase atmospheric Δ14C to glacial levels, a drastic reduction of air-sea exchange efficiency in the polar regions must be assumed, though discrepancies remain for the portion of the record younger than ~ 33 kyr BP. We end with an illustration of how the 14C production rate would have had to evolve to be consistent with the Δ14C record, by combining an atmospheric radiocarbon budget with the Bern3D model. The overall conclusion is that the remaining discrepancies with respect to glacial Δ14C may be linked to an underestimation of 14C production and/or a biased-high reconstruction of Δ14C over the time period of interest. Alternatively, we appear to still be missing an important carbon cycle process for atmospheric Δ14C.