We describe the performance of the Fermi LAT instrument in the context of the Pass 7 event selection. Pass 7 was developed by the LAT team based on the experience gained in the first years of operation and was released to the public in August 2011. We discuss the on-orbit effects that motivated updating the pre-launch data reduction process and describe the corrections that we apply to the Monte Carlo-derived Instrument Response Functions (IRFs) to account for the discrepancies observed between flight and simulated data. We also give details of the validations performed using flight data and quantify the residual uncertainties in the IRFs. Finally, we describe techniques that the LAT team has developed to propagate those uncertainties into estimates of the systematic errors on common measurements such as fluxes and spectra of astrophysical sources.