Carbon dioxide is the most abundant volatile in kimberlite melts and its solubility exerts a prime influence on the melt structure, buoyancy, transport rate and hence eruption dynamics. The actual primary composition of kimberlite magma is the matter of some debate but the solubility of CO2 in kimberlitic melts is also poorly constrained due to difficulties in quenching these compositions to a glass that retains the equilibrium CO2 content. In this study we used a range of synthetic, melt compositions with broadly kimberlitic to carbonatitic characteristics which can, under certain conditions, be quenched fast enough to produce a glass. These materials are used to determine the CO2 solubility as a function of chemical composition and pressure (0.05–1.5 GPa). Our results suggest that the solubility of CO2 decreases steadily with increasing amount of network forming cations from ~ 30 wt.% CO2 at 12 wt.% SiO2 down to ~ 3 wt.% CO2 at 40 wt.% SiO2. For low silica melts, CO2 solubility correlates non-linearly with pressure showing a sudden increase from 0.1 to 100 MPa and a smooth increase for pressure > 100 MPa. This peculiar pressure–solubility relationship in low silica melts implies that CO2 degassing must mostly occur within the last 3 km of ascent to the surface having potential links with the highly explosive nature of kimberlite magmas and some of the geo-morphological features of their root zone. We present an empirical CO2 solubility model covering a large range of melt composition from 11 to 55 wt.% SiO2 spanning the transition from carbonatitic to kimberlitic at pressures from 1500 to 50 MPa.