The Gaia DR2 radial velocity data includes median radial velocities for 7.22 million stars with effective temperatures between 3550 and 6900 K. These velocities were validated using ground-based catalogues and show a completeness of 77.2% for stars with G ≤ 12.5 mag. The median radial velocity residuals with respect to ground-based surveys vary between catalogues but do not exceed a few hundred m/s. The radial velocities show a positive trend with magnitude, reaching +500 m/s at G_RVS ≈ 11.75 mag. The overall precision is 1.05 km/s, with higher precision for bright stars (200-350 m/s) and lower precision for faint stars (1.4-3.7 km/s). The radial velocity precision depends on magnitude and effective temperature. The data were processed using the RVS instrument, which has a resolving power of R = 11,500 and covers wavelengths from 845 to 872 nm. The pipeline includes validation steps to select high-quality radial velocities, including filters for large coordinate uncertainties, faint stars, ambiguous transits, large radial velocity uncertainties, suspected double-line binaries, suspected emission line stars, cool stars, hot stars, and high-velocity stars. The final radial velocity catalogue includes median velocities, uncertainties, number of transits, and template parameters. The accuracy of the radial velocities was assessed using five ground-based catalogues, showing a small offset with respect to CU6GB and a positive offset with respect to SIM, RAVE, APOGEE, and GES. The accuracy varies with effective temperature, surface gravity, and metallicity, with a positive trend with metallicity and a decrease with surface gravity. The accuracy also shows a positive trend with magnitude, which is attributed to Charge Transfer Inefficiency (CTI) effects. The radial velocity data is validated and published with a full sky coverage and high completeness for bright stars.The Gaia DR2 radial velocity data includes median radial velocities for 7.22 million stars with effective temperatures between 3550 and 6900 K. These velocities were validated using ground-based catalogues and show a completeness of 77.2% for stars with G ≤ 12.5 mag. The median radial velocity residuals with respect to ground-based surveys vary between catalogues but do not exceed a few hundred m/s. The radial velocities show a positive trend with magnitude, reaching +500 m/s at G_RVS ≈ 11.75 mag. The overall precision is 1.05 km/s, with higher precision for bright stars (200-350 m/s) and lower precision for faint stars (1.4-3.7 km/s). The radial velocity precision depends on magnitude and effective temperature. The data were processed using the RVS instrument, which has a resolving power of R = 11,500 and covers wavelengths from 845 to 872 nm. The pipeline includes validation steps to select high-quality radial velocities, including filters for large coordinate uncertainties, faint stars, ambiguous transits, large radial velocity uncertainties, suspected double-line binaries, suspected emission line stars, cool stars, hot stars, and high-velocity stars. The final radial velocity catalogue includes median velocities, uncertainties, number of transits, and template parameters. The accuracy of the radial velocities was assessed using five ground-based catalogues, showing a small offset with respect to CU6GB and a positive offset with respect to SIM, RAVE, APOGEE, and GES. The accuracy varies with effective temperature, surface gravity, and metallicity, with a positive trend with metallicity and a decrease with surface gravity. The accuracy also shows a positive trend with magnitude, which is attributed to Charge Transfer Inefficiency (CTI) effects. The radial velocity data is validated and published with a full sky coverage and high completeness for bright stars.