During its tour of the Saturn system, Cassini performed two close flybys of Rhea dedicated to gravity investigations, the first in November 2005 and the second in March 2013. This paper presents an estimation of Rhea’s fully unconstrained quadrupole gravity field obtained from a joint multi-arc analysis of the two Cassini flybys. Our best estimates of the main gravity quadrupole unnormalized coefficients are J2 10^6 = 94 6.0 ± 13.9, C22 10^6 = 242.1 ± 4.0 (uncertainties are 1-r). Their resulting ratio is J2/C22 = 3.91 ± 0.10, statistically not compatible (at a 5-r level) with the theoretical value of 10/3, predicted for a hydrostatic satellite in slow, synchronous rotation around a planet. Therefore, it is not possible to infer the moment of inertia factor directly using the Radau–Darwin approximation. The observed excess J2 (gravity oblateness) was investigated using a combined analysis of gravity and topography, under different plausible geophysical assumptions. The observed gravity is consistent with that generated by the observed shape for an undifferentiated (uniform density) body. However, because the surface is more likely to be water ice, a two-layer model may be a better approximation. In this case, and assuming a mantle density of 920 kg/m3, some 1–3 km of excess core oblateness is consistent with the observed gravity. A wide range of moments of inertia is allowed, but models with low moments of inertia (i.e., more differentiation) require greater magnitudes of excess core topography to satisfy the observations.