The impact-parameter dependent color glass condensate (b-CGC) dipole model is based on the Balitsky-Kovchegov nonlinear evolution equation and improves the Iancu-Itakura-Munier dipole model by incorporating the impact-parameter dependence of the saturation scale. Here we confront the model to the recently released high precision combined Hadron Electron Ring Accelerator (HERA) data and obtain its parameters. The b-CGC results are then compared to data at small x for the structure function, the longitudinal structure function, the charm structure function, exclusive vector meson (J/psi, phi, and rho) production and deeply virtual Compton scattering. We also compare our results with the impact-parameter dependent saturation (IP-Sat) model. We show that most features of inclusive deep inelastic scattering and exclusive diffractive data, including the Q(2), W, vertical bar t vertical bar, and x dependence, are correctly reproduced in both models. Nevertheless, the b-CGC and the impact-parameter dependent saturation (IP-Sat) models give different predictions beyond the current HERA kinematics, namely for the structure functions at very low x and high virtualities Q(2), and for the exclusive diffractive vector meson and deeply virtual Compton scattering production at high t. This can be traced back to the different power-law behavior of the saturation scale in x and to a different impact-parameter b dependence of the saturation scale in these models. Nevertheless, both models give approximately similar saturation scales Q(S) < 1 GeV for the proton in HERA kinematics, and also both models lead to the same conclusion that the typical impact parameter probed in the total gamma* p cross section is about b approximate to 2-3 GeV-1. Our results provide a benchmark for further investigation of QCD at small x in heavy ion collisions at RHIC and the LHC and also at future experiments such as an electron-ion collider and the LHeC.