In spite of its major role on the atmospheric volatile budget, climate, and tracking magmatic transfers, mantle (CO2) degassing below volcanoes is still poorly understood. Most of the studies on this scientific topic lack constraint on the CO2 concentration of primary melts, the depth at which it starts degassing, and the extent of this process in the mantle. In this study of Piton de la Fournaise (PdF) volcano, we couple geochemistry of low solubility gases (He, Ar, CO2, δ13C) in fluid inclusions (FIs) and petro-chemistry of magmatic inclusions on a set of olivine and clinopyroxene crystals from basalts and ultramafic enclaves. We constrain basaltic melt degassing at PdF over a large pressure range (from 4 GPa up to the surface). Based on CO2-He-Ar systematics, we infer that extensive degassing occurs already in the upper mantle (4–1 GPa) and it is favored by multiple steps of magma ponding and differentiation up to the mantle-crust underplating depth (0.4 GPa). Thus, we calculate that basaltic melts injected at crustal depth (<0.4 GPa) have already exsolved ∼94 ± 5 wt% of their primary CO2 content in accordance with (1) the evolved and degassed signature of erupted lavas and (2) the weakness of inter-eruptive gas emissions in the active area bearing low-temperature vapor-dominated fumaroles. Our results at PdF strongly contrast with previous findings on other ocean island volcanoes having a higher magma production rate and faster magma ascent, like Kilauea (Hawaii), whose basalts experience only limited extent of differentiation and degassing. We propose that extensive degassing already in the upper mantle can be a common process for many volcanoes of the Earth and is tightly dependent on the dynamics of magma ascent and differentiation across multiple ponding zones. Based on the modeling developed in this study, we propose a new estimation of the CO2 content (up to 3.5 ± 1.4 wt%) in primary basaltic melts at PdF leading to a carbon content in the mantle source of 716 ± 525 ppm. This new estimation is considerably higher than the few previous calculations performed for Ocean Island Basalts (OIB) systems. Another implication of this work involves the possible bias between the δ13C measured in volcanic gas emissions (<−6‰) and that of primary vapour phase (−0.5 ± 0.5‰) constrained in this work. This bias would confirm the early step of extensive CO2 degassing within the upper mantle and could represent an alternative for the hypotheses of carbon recycling or mantle heterogeneity in support of the low δ13C signature of some mantle reservoirs. This study bears significant implications on the global budget of volcanic volatile emissions, chiefly regarding the contribution of past and future emissions of volcanic CO2 to climate dynamics, and on volcanic gas monitoring.