The dynamical origin of the inner solar system remains heavily debated. Models are constrained by the masses and orbits of the terrestrial planets, the very low mass and structure of the asteroid belt, and the architecture of the outer solar system. It is well-accepted that the late stage of terrestrial accretion is characterized by giant collisions among Mars-mass planetary "embryos". Two contrasting views exist regarding the state of the inner system at this stage. The first view assumes a smooth mass distribution of terrestrial bodies with up to a few Earth masses in solids in the asteroid region. To match the current belt in this view, models must deplete the asteroid region by a factor of 100-1000. The classical model systematically fails to reproduce the large Earth/Mars mass ratio and and tends to strand Mars-mass embryos in the belt, which is inconsistent with the present-day belt. New models have been built taking into account the fact that matching Mars' small mass requires a severe mass deficit beyond 1 au. The Grand-Tack model invokes a specific migration history of the giant planets to deplete Mars' feeding zone, to deplete and excite the asteroid belt, successfully matching constraints. Yet uncertainties in the giant planets' growth and migration histories have motivated alternative scenarios that do not require large scale radial migration. Another viable model invokes also an early dynamical instability in the giant planet's orbits to deplete the asteroid belt and Mars' feeding zone. In the second, contrasting view the asteroid belt region was already strongly mass depleted before the late phase of terrestrial accretion. The low-mass asteroid belt model starts with a belt a few times more massive than the current one and naturally produces good Mars analogs. The giant planets can subsequently excite the asteroids' orbits to the current observed levels in a way that is consistent with models of the dynamical evolution of the outer solar system. In this talk we will review these scenarios and discuss their implications and potential ways to distinguish them.< Réduire