The 2025 \(\:{M}_{w}\) 7.0 Xizang Dingri normal-fault earthquake offers a rare opportunity to evaluate the performance of existing ground-motion models (GMMs) for large normal-fault events ( \(\:{M}_{w}\) ≥ 7.0). Using 35 three-component strong-motion records spanning 35–300 km, we assessed four NGA-West2 models and two Chinese regional models (CEA2019, ZB2022). Results show that while NGA-West2 models perform well for low-frequency ground motion parameters (e.g., PSA (T ≥ 3 s)), they systematically overpredict mid- to high-frequency parameters (PGA, PSA (T < 2.5 s)) in both near- and far-field, likely due to differences in source characteristics and strong crustal attenuation in the Qinghai-Xizang Plateau. The ZB2022 model, although calibrated using data from the eastern Qinghai–Xizang Plateau, also exhibits high-frequency (PSA (T < 0.5 s)) overestimation, possibly due to magnitude extrapolation limitations and regional heterogeneity. The CEA2019 model performs well in the near-field but significantly underpredicts long-period motions in the far-field, which may be related to limitations in its modeling methodology. Duration analysis shows both AS16 and WEN18 models underestimate long-duration shaking at far-field. These discrepancies highlight the limitations of existing GMMs in capturing ground motions of large normal-fault earthquakes in Qinghai–Xizang plateau. Our findings emphasize the need for updated regional GMMs incorporating plateau-specific source and attenuation characteristics, particularly for improving ground motion predictions in future seismic hazard assessments.
