Purpose <p>As obesity has reached pandemic proportions worldwide, improving technical solutions for its treatment requires robust planning and numerical modelling. Yet existing material models obtained from biaxial tensile test are scarce and based only on frozen human or porcine samples, without histological quantitative consideration. Our objective was to generate fresh-human biaxial data including microstructure orientation/dispersion and provide patient-specific material models suitable for biofidelic FE modelling.</p> Methods <p>Fundus and corpus samples of human stomach (10 sleeve-gastrectomy patients) underwent planar biaxial testing at two rates (0.1 and 1&#xa0;mm·s⁻¹) and three displacement ratios (1:1, 1:2, 2:1). Full-field strain was measured by digital image correlation. Collagen orientation/dispersion were extracted from histology using machine learning techniques. An incompressible anisotropic hyperelastic law (neo-Hookean matrix + two fibre families) was fitted per specimen across all ratios/directions.</p> Results <p>A consistent hierarchy emerged despite inter-patient variability with corpus stiffer than fundus, and longitudinal orientation stiffer than circumferential. Higher speed resulted in higher stress. In terms of safe characterisation limit (measured at 0.1&#xa0;mm·s⁻¹), median true-strain at safe characterisation limit was 0.33 (corpus) vs 0.38 (fundus). Isotropic matrix-only model failed to fit ratios/directions simultaneously while anisotropic model reproduced multi-ratio responses with R² &gt; 0.7 both in circumferential and longitudinal in respectively 85 and 81% of specimen.</p> Conclusion <p>Fresh-human, full-field biaxial data coupled to quantified collagen architecture yield specimen-level parameters and region-specific safe-stretch thresholds, enabling biofidelic, patient-specific FE simulations of gastric procedures and device–tissue interaction.</p>