Meter-scale large-aperture gratings are essential in petawatt-class picosecond laser systems. Their grating mounts must support heavy-load arrays and high alignment accuracy due to high energy density and long beam paths. However, nonlinear errors from parasitic motions and transmission gaps can significantly degrade precision.This study presents a kinetostatics modeling and error calibration framework for the grating mount, incorporating an improved particle swarm optimization (PSO) algorithm. The nonlinear error model combines energy-based and pseudo-rigid-body methods, with equivalent representations of structural gaps and parasitic motions. To capture multi-source nonlinear interactions, a global–dynamic multi-subgroup PSO enhances calibration via coordinated global exploration and local refinement. Experiments indicate, compared with conventional models, first-round compensation reduces average errors by over 65.4%, 79.8%, and 74.8% in rotation, tip, and tilt, respectively. The method integrates nonlinear pose modeling, unified gap representation, and an enhanced PSO strategy, offering an effective solution for error compensation in meter-scale, heavy-load compliant mechanisms.