Physical reasoning is a core aspect of intelligence in animals and humans. A\ncentral question is what model should be used as a basis for reasoning.\nExisting work considered models ranging from intuitive physics and physical\nsimulators to contact dynamics models used in robotic manipulation and\nlocomotion. In this work we propose descriptions of physics which directly\nallow us to leverage optimization methods for physical reasoning and sequential\nmanipulation planning. The proposed multi-physics formulation enables the\nsolver to mix various levels of abstraction and simplifications for different\nobjects and phases of the solution. As an essential ingredient, we propose a\nspecific parameterization of wrench exchange between object surfaces in a path\noptimization framework, introducing the point-of-attack as decision variable.\nWe demonstrate the approach on various robot manipulation planning problems,\nsuch as grasping a stick in order to push or lift another object to a target,\nshifting and grasping a book from a shelve, and throwing an object to bounce\ntowards a target.\n