Recently, throwing is gaining even more research interest, especially in logistics, where goods have to be moved efficiently. The throwing approach is difficult because it requires precision in a dynamic task. In this work, we consider the problem in both joint and cartesian space. In joint space, we use an optimization procedure to obtain a trajectory to follow, maximizing the throwing distance. In task space, we use a cartesian controller to throw in a desired position. In both cases, minimum-jerk trajectories are used to have smooth signals and deal with robot constraints, usually present in collaborative robots. The methods are tested on simulations and experiments on the 7 d.o.f Franka Emika Panda.