We numerically investigate the Potential Energy Landscape of an off-lattice beta-sheet model protein, looking at saddles and minima probed by the system during the folding process. Go-like (with native-state-dependent force field and funnel-like landscape) and non-Go-like models are considered. In the Go-like case, on varying the temperature, we observe: i) a pronounced peak at the collapse/folding temperature T(theta) similar or equal to T(f) in the energy elevation of visited saddles from underlying minima, ii) a crossover at the same point of the saddle order. Saddles-based quantities seem then to be good candidates as indicators of the funneled shape of the landscape in protein models. Copyright (C) EPLA, 2009