Abstract Melt inclusions represent a unique tool to reconstruct the composition and chemical evolution of silicate melts in magmatic systems. Laser-ablation inductively-coupled-plasma mass-spectrometry (LA-ICP-MS) is the most commonly used microanalytical technique to analyse crystallized melt inclusions without prior re-homogenization. Well-preserved melt inclusions can be quantified by subtracting the contribution of co-ablated host with a carefully selected internal standard. However, post-entrapment compositional re-equilibration commonly renders this task difficult, to the same degree as it would affect any quantification after prior re-homogenization. In this study, we first examine well-preserved, crystallized melt inclusions hosted in olivine, plagioclase, apatite, clinopyroxene and orthopyroxene from porphyry dikes and volcanic rocks to test various quantification strategies and evaluate the associated uncertainties, and then we use these strategies to quantify coarsely crystallized melt inclusions from gabbroic rocks at Marble Canyon (USA) and Laiyuan (China) that experienced severe post-entrapment modifications due to relatively slow cooling rates. The results demonstrate that even for well-preserved melt inclusions hosted in chemically complex minerals the uncertainty related to inclusion–host deconvolution can be rather high (up to 30‒50% for host-incompatible trace elements significantly above their limits of detection), though other uncertainties inherent to LA-ICP-MS analysis are relatively small (typically ≤5‒10%). The deconvolution-related uncertainty can be minimized to ca. 10% by (1) choosing whole rocks that are fresh and representative of magmatic liquids, (2) choosing the smallest possible spot size to ablate the melt inclusions, and (3) choosing a host endmember that is compositionally as similar as possible to the one ablated together with the inclusion. Results of coarsely crystallized melt inclusions from gabbroic rocks suggest that the range of elements affected by post-entrapment re-equilibration varies from intrusion to intrusion. Olivine-hosted melt inclusions from Marble Canyon appear to have diffusively lost Fe, Ti and Ca, whereas those from Laiyuan lost Fe, Na, Al, Ca, Ti and Y and gained V. However, the relative abundances of K, P, Rb, Sr, Zr, Nb, Mo, Cs, Ba, Ce, Ta, Pb, Th, U and ±Cu appear unchanged. Plagioclase-hosted melt inclusions from Marble Canyon are relatively well-preserved, whereas those from Laiyuan lost significant amounts of Fe, K, Mg, Mn, Rb and Co. Apatite-hosted melt inclusions seem well preserved with regard to most elements except for Cu. These results suggest that despite the post-entrapment modification of certain element concentrations and the associated difficulties in melt inclusion quantification, information on the approximate abundances of other elements that are invaluable for petrogenetic and metallogenic studies can still be retrieved from melt inclusions in gabbroic rocks using the LA-ICP-MS technique.