The evolution of the frictional strength along a fault at seismic slip rates (about 1 m/s) is a key factor controlling earthquake mechanics. At mantle depths, friction‐induced melting and melt lubrication may influence earthquake slip and seismological data. We report on laboratory experiments designed to investigate dynamic fault strength and frictional melting processes in mantle rocks. We performed 20 experiments with Balmuccia peridotite in a high‐velocity rotary shear apparatus and cylindrical samples (21.8 mm in diameter) over a wide range of normal stresses (5.4–16.1 MPa), slip rates (0.23–1.14 m/s), and displacements (1.5–71 m). During the experiments, shear stress evolved with cumulative displacement in five main stages (stages 1–5). In stage 1 (first strengthening), the coefficient of friction μ increased up to 0.4–0.7 (first peak in friction). In stage 2 (abrupt first weakening), μ decreased to about 0.25–0.40. In stage 3 (gradual second strengthening), shear stress increased toward a second peak in friction (μ = 0.30–0.40). In stage 4 (gradual second weakening), the shear stress decreased toward a steady state value (stage 5) with μ = 0.15. Stages 1 and 2 are of too short duration to be investigated in detail with the current experimental configuration. By interrupting the experiments during stages 3, 4, and 5, microstructural (Field Emission Scanning Electron Microscope) and geochemical (Electron Probe Micro‐Analyzer and Energy Dispersive X‐Ray Spectroscopy) analysis of the slipping zone suggest that second strengthening (stage 3) is associated with the production of a grain‐supported melt‐poor layer, while second weakening (stage 4) and steady state (stage 5) are associated with the formation of a continuous melt‐rich layer with an estimated temperature up to 1780°C. Microstructures formed during the experiments were very similar to those found in natural ultramafic pseudotachylytes. By performing experiments at different normal stresses and slip rates, (1) the “thermal” (as it includes the thermally activated first and second weakening) slip distance to achieve steady state from the first peak in strength decreased with increasing normal stress and slip rate and (2) the steady state shear stress slightly increased with increasing normal stress and, for a given normal stress, decreased with increasing slip rate. The ratio of shear stress versus normal stress was about 0.15, well below the typical friction coefficient of rocks (0.6–0.8). The dependence of steady state shear stress with normal stress was described by means of a constitutive equation for melt lubrication. The presence of microstructures similar to those found in natural pseudotachylytes and the determination of a constitutive equation that describes the experimental data allows extrapolation of the experimental observations to natural conditions and to the study of rupture dynamics in mantle rocks.