Due to the significant increase in the number of monitor units used to deliver a dynamic IMRT treatment, the total MLC leakage (transmission plus scatter) can exceed 10% of the maximum in-field dose. To avoid dosimetric errors, this leakage must be accurately accounted for in the dose calculation and conversion of optimized intensity patterns to MLC trajectories used for treatment delivery. In this study, we characterized the leaf end transmission and leakage radiation for Varian 80- and 120-leaf MLCs using Monte Carlo simulations. The complex geometry of the MLC, including the rounded leaf end, leaf edges (tongue-and-groove and offset notch), mounting slots, and holes was modeled using MCNP4b. Studies were undertaken to determine the leakage as a function of field size, components of the leakage, electron contamination, beam hardening and leaf tip effects. The leakage radiation with the MLC configured to fully block the field was determined. Dose for 6 and 18 MV beams was calculated at 5 cm depth in a water phantom located at 95 cm SSD, and normalized to the dose for an open field. Dose components were scored separately for radiation transmitted through and scattered from the MLC. For the 80-leaf MLC at 6 MV, the average leakage dose is 1.6%, 1.7%, 1.8%, and 1.9% for 5 x 5, 10 x 10, 15 x 15, and 20 x 20cm2 fields, respectively. For the 120-leaf MLC at 6 MV, the average leakage dose is 1.6%, 1.6%, 1.7%, and 1.9% for the same field sizes. Measured leakage values for the 120-leaf MLC agreed with calculated values to within 0.1% of the open field dose. The increased leakage with field size is attributed to MLC scattered radiation. The fractional electron contamination for a blocked MLC field is greater than that for an open field. The MLC attenuation significantly affects the photon spectrum, resulting in an increase in percent depth dose at 6 MV, however, little effect is observed at 18 MV. Both phantom scatter and the finite source size contribute to the leaf tip profile observed in phantom. The results of this paper can be applied to fluence-to-trajectory and trajectory-to-fluence calculations for IMRT.