The mass of volatiles degassed from volcanic eruptions is often estimated by comparing the volatile concentrations in undegassed glassy melt inclusions with the volatile concentrations in the degassed matrix glass. However, melt inclusions are prone to post-entrapment modification, including diffusive H+ loss through the host olivine crystal lattice which lowers the H2O content of the inclusion, and the degassing of CO2 into a bubble in response to cooling and crystallisation on the inclusion walls. Such bubbles are very common in olivine-hosted melt inclusions from the AD 1783–1784 Laki eruption, south–east Iceland. We have determined the CO2 content of these bubbles using micro-Raman spectroscopy, and the CO2 concentration in the glass by SIMS. Our results show that >90% of the total inclusion CO2 may be sequestered into the bubble, which demonstrates the importance of measuring the compositions of both vapour bubbles and the glass phase in melt inclusions. We reconstruct the deep degassing path of the Laki magma by using Nb as proxy for the undegassed CO2 content of the melt inclusions. The substantial CO2/Nb variation in the Laki melt inclusions (3.8–364) can be explained by concurrent crystallisation and CO2 degassing in the Laki magmatic system. We calculate the amount of CO2 lost from individual melt inclusions, assuming CO2/Nb ≈ 435 for enriched Icelandic mantle and CO2/Nb ≈ 171 for depleted mantle. Melt inclusions with the greatest saturation pressures have lost the least CO2 prior to inclusion trapping. At any given saturation pressure, the most enriched melt inclusions have lost the most CO2, while the most depleted inclusions have lost very little CO2. Enriched primary melts with high initial CO2 concentrations are therefore useful for investigating deep degassing behaviour in magmatic systems because a range of melt inclusion saturation pressures are recorded during crystallisation and degassing. Depleted melt inclusions with low initial CO2 concentrations remain vapour-undersaturated to shallow levels and cannot be used to constrain deep degassing behaviour. The cumulative CO2 mass release from the Laki magma is determined as a function of pressure and extent of crystallisation. Using an updated petrologic method that takes into account the diversity of primary melts and CO2 sequestration into vapour bubbles, we calculate the total mass of CO2 exsolved from the Laki magma to be 304 Mt.