To address the limited performance of global zenith tropospheric delay (ZTD) models in representing strong elevation effects and local climatic variability over complex terrain in Xizang, a 0.25° × 0.25° high-resolution gridded ZTD model (XZZM) is developed by fusing GNSS-derived ZTD observations from 49 stations (2017–2019) with contemporaneous European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis data. The model captures pronounced annual and seasonal ZTD variability and incorporates a vertical-profile formulation to improve near-surface refractivity representation. The elevation normalization parameter β exhibits clear spatial and temporal heterogeneity, with its residuals increasing approximately linearly with elevation normalization error. Validation at 49 stations indicates an overall ZTD accuracy of approximately 10–20 mm, with smaller RMSE and more stable seasonal performance than ERA5. Compared with the Global Pressure and Temperature 3 (GPT3) model, XZZM exhibits more concentrated residuals, improved seasonal consistency, and higher overall statistical accuracy. Precise point positioning (PPP) tests at 12 sites further demonstrate that XZZM improves vertical positioning accuracy by 35.9% relative to GPT3. These results indicate that XZZM provides reliable tropospheric delay corrections for GNSS positioning and atmospheric applications in complex plateau environments.