This study aims to identify optimal sampling designs for the inventory and monitoring of medium- and large-sized mammals, considering different biodiversity dimensions (taxonomic, functional, and phylogenetic) in the Southern Brazilian Amazon. We established three line transects each 3 km long, and three camera trap grids, each with six cameras, totalling 176 kilometers walked and 4,914 camera trap-days. We defined fixed and variable costs associated with each method. We sought to identify the sampling arrangement that yielded the highest possible α-diversity at the lowest possible cost (i.e., inventory efficiency), and the combination of methods that minimized bias in recording β-diversity while also minimizing costs (i.e., monitoring efficiency). Camera traps detected 26 species, of which 16 were exclusive to this method. Line transects resulted in detection of 16 species, six of which were exclusive (all arboreal). It was generally not possible to identify a single sampling scheme that yielded higher diversity or lower bias at lower costs. However, it was clear that adding line transect sampling units increased costs without improving diversity or bias results. Then, for the inventory of functional and phylogenetic diversity and for monitoring taxonomic, functional, and phylogenetic diversity, the optimal sampling design involves the exclusive use of camera traps. For a taxonomic diversity inventory the optimal sampling scheme requires a combination of camera traps and line transects. We did not sweep the transects or search for tracks and other signs in our line transect surveys, which reduced the method’s effectiveness in detecting some species. The superior cost-effectiveness of camera traps can be related to improvements in camera trap technology and reductions in their cost, although given their lower efficiency to detect arboreal species, including some effort in line transects may be necessary.