Mechanical metamaterials enable programmable structural responses through geometry rather than material composition. In this work, we present strain-programmable strut-based lattice metamaterials capable of exhibiting multiple elastic phases, each with a distinct effective modulus, within a single material system. The proposed lattices combine full-length load-bearing struts with fractional struts that remain inactive at low strains and progressively engage as deformation increases, producing discrete, strain-driven stiffness transitions. This mechanism allows precise control of phase-dependent elasticity without relying on material nonlinearity, buckling, or multi-material fabrication. A computational mechanics framework is developed to model the phase-dependent elastic response of these lattices. Finite-element simulations, complemented by systematic parametric studies, identify the key geometrical parameters governing stiffness evolution and phase transitions. Machine learning–assisted inverse design is employed to efficiently generate lattice topologies that achieve prescribed multi-phase elastic behavior. Additively manufactured specimens validate the numerical predictions under compressive loading. The proposed multi-phase metamaterials provide a robust platform for mechanically programmable architectures with tailored, phase-dependent elastic properties, opening avenues for adaptive structures, programmable mechanical systems and compliant mechanisms.