In this paper we report, clarify and broaden various recent efforts to complement the chemistry-centered models of force generation in muscles by mechanics-centered models. The physical mechanisms of interest can be grouped into two classes: passive and active. The main passive effect is the fast force recovery which does not require the detachment of myosin cross-bridges from actin filaments and can operate without a specialized supply ATP. In mechanical terms, it can be viewed as a collective folding-unfolding phenomenon in the system of interacting bistable units and modeled by near equilibrium Langevin dynamics. The parallel active force generation mechanism operates at slow timescales, requires detachment and is crucially dependent on ATP hydrolysis. The underlying mechanical processes take place far from equilibrium and are represented by stochastic models with broken time reversal symmetry implying non-potentiality, correlated noise or multiple reservoirs. The modeling approaches reviewed in this paper deal with both active and passive processes and support from the mechanical perspective the biological point of view that phenomena involved in slow (active) and fast (passive) force generation are tightly intertwined. They reveal, however, that biochemical studies in solution, macroscopic physiological measurements and structural analysis do not provide by themselves all the necessary insights into the functioning of the contractile system. In particular, the reviewed body of work emphasizes the important role of long-range interactions and criticality in securing the targeted mechanical response in the physiological regime of isometric contractions. The importance of the purely mechanical microscale modeling is accentuated at the end of the paper where we address the puzzling issue of the stability of muscle response on the so called descending limb of the isometric tetanus.