<p>This paper examines the shape memory effect (SME) of PLA-TPU polymer blends that have been produced via a novel method of pellet-based additive manufacturing called Fused Granular Fabrication (FGF). The melt-blending of PLA with TPU (Flex 93A) in three ratios (10, 20, and 30 wt%) was performed through a single-screw extruder, followed by the pelletization and 3D printing into cuboidal shape specimens with different infill patterns (honeycomb, grid, triangular) and densities (10–40%). Mechanical programming of the samples was conducted through the application of compressive strains (10%, 25% and 40%), and the samples were thermally activated to study recovery behavior. A Taguchi L27 orthogonal array, ANOVA, and Grey Relational Analysis (GRA) were used to analyze four important response variables: shape fixity (R_f), shape recovery (R_r), overall shape memory effect (SME), and peak compressive strength (PCS). Findings revealed that the programming strain was the strongest parameter with a maximum contribution of 70.71% to variation in recovery. The ratio of TPU in the blend and infill determined the elasticity and mechanical performance, respectively. The best combination was determined to consist of 10 wt% TPU, 40% infill density, honeycomb design, and 40% programming strain. In general, the presented results demonstrate the promise of PLA-TPU blends produced through FGF in the 4D-printed thermally responsive and reconfigurable structures to be used in practice in biomedical applications, deployable systems, and soft robotics.</p>