Purpose: To evaluate the physical performance of Megavoltage Cone‐Beam CT (MVCBCT) and to optimize system and reconstruction settings for image quality. Methods and Materials: Several system parameters were varied to quantify their impact on image quality including the exposure (2.7, 4.5, 9.0, 18.0 and 54.0 MU), the cranio‐caudal field‐size (2, 5, 15, 27.4 cm), the voxel size (0.5, 1, 2 mm)and the slice thickness (1, 3, 5 mm). For the reconstruction algorithm, we investigated binning, averaging and diffusion of raw projections as well as four different backprojection filters. Two CT♯ normalization factors were compared. A head size water cylinder with different configurations of CT inserts was used to measure contrast‐to‐noise ratio (CNR) and uniformity. The point‐spread function (PSF) was obtained using a brass wire and an iterative edge blurring technique. The current MVCBCT product settings were used as the performance baseline for comparison. Results: Beam intensity variations per projection of up to 35.4% were observed for a 2.7 MU MVCBCT acquisition. Such variations were mostly captured in the system MU reading per frame and did not affect the CNR. The non‐uniformity was reduced from 18.8% to 14.2% by closing the Y‐jaws for imaging. An optimized reconstruction protocol was developed and showed an improvement of 60% in CNR with a penalty of only 8%for the PSF and an increase of 1 to 2 minutes in reconstruction time. The application of diffusion filtering for 9 MU reconstructions resulted in similar CNR improvement to using 5 times more dose with the current reconstruction protocol. Using reconstructions with smaller voxels and thicker slices can further improve the CNR.Conclusion: The image quality stability of MVCBCT over a 4‐month period was excellent. Soft‐tissue visualization with MVCBCT can be substantially improved with proper system settings. Conflict of Interest: Research sponsored by Siemens OCS.