The paper summarizes the third Monocular Depth Estimation Challenge, which tested algorithms on their ability to generalize to complex natural and indoor scenes. 19 submissions outperformed the baseline method. 10 teams submitted reports, showing widespread use of Depth Anything model. The top method increased the 3D F-Score from 17.51% to 23.72%.