Here I will present an overview of what we have learned about ongoing and future changes in precipitation extremes from analyses of the CMIP6 ensemble. I will review our understanding of the ability of GCMs to simulate extreme precipitation, including how model quality is affected by resolution and parameterization choices, and how biases in local atmospheric dynamics affect extreme precipitation. I will also discuss the simulated changes in extreme precipitation induced by anthropogenic forcing in the CMIP6 ensemble, and whether such changes are already detectable in the observational record.