Animals can amplify the mechanical power output of their muscles as they jump to escape predators or strike to capture prey. One mechanism for amplification involves muscle-tendon (MT) systems in which a spring element is pre-stretched while held in place by a ‘latch’ that prevents immediate transmission of muscle power to the load. In principle, this storage phase is followed by a triggered release of the ‘latch’, and elastic energy released from the spring element enables muscle to exceed its maximum power limit (Pamp=Pmt/Pmax muscle >1.0). Latches enable power amplification by increasing the muscle work generated during storage and reducing the duration over which that stored energy is released to power a movement. Previously described biological ‘latches’ include: skeletal levers, anatomical triggers, accessory appendages and even antagonist muscles. In fact, many species that rely on high-powered movements also have a large number of muscles arranged in antagonist pairs. Here, we examine whether antagonist muscles can be useful as active latches to achieve controlled power amplification. We developed a computer model of a frog hindlimb driven by a compliant MT. We simulated MT power generated against an inertial load in the presence of an antagonist muscle ‘latch’ (AML) with relaxation time varying from very fast (10ms) to very slow (1000ms). The fastest AML produced power amplification (Pamp =5.0) while the slowest AML produced power attenuation (Pamp =0.43). Notably, AMLs with relaxation times shorter than ~300ms also yielded greater power amplification (Pamp >1.20) than the system driving the same inertial load using only an agonist MT without any AML. Thus, animals that utilize a sufficiently fast relaxing AML ought to be capable of achieving greater power output than systems confined to a single agonist MT tuned for maximum Pamp against the same load.