We use gravity implied by the Earth's rock-equivalent topography (RET) and modeled isostatic compensation masses to evaluate the new global gravity field models (GGMs) from European Space Agency (ESA)'s Gravity Field and Steady-State Ocean Circulation Explorer (GOCE) satellite gravimetry mission. The topography is now reasonably well-known over most of the Earth's landmasses, and also where conventional GGM evaluation is prohibitive due to the lack (or unavailability) of ground-truth gravity data. We construct a spherical harmonic representation of Earth's RET to derive band-limited topography-implied gravity, and test the somewhat simplistic Airy/Heiskanen and Pratt/Hayford hypotheses of isostatic compensation, but which did not improve the agreement between gravity from the uncompensated RET and GOCE. The third-generation GOCE GGMs (based on 12 months of space gravimetry) resolve the Earth's gravity field effectively up to spherical harmonic degree ∼200-220 (∼90-100 km resolution). Such scales could not be resolved from satellites before GOCE. From the three different GOCE processing philosophies currently in use by ESA, the time-wise and direct approaches exhibit the highest sensitivity to short-scale gravity recovery, being better than the space-wise approach. Our topography-implied gravity comparisons bring evidence of improvements from GOCE to gravity field knowledge over the Himalayas, Africa, the Andes, Papua New Guinea and Antarctic regions. In attenuated form, GOCE captures topography-implied gravity signals up to degree ∼250 (∼80 km resolution), suggesting that other signals (originating, e.g., from the crust-mantle boundary and buried loads) are captured as well, which might now improve our knowledge on the Earth's lithosphere structure at previously unresolved spatial scales.