The thermal regime is a key indicator of permafrost evolution and thaw trajectories in response to climate change, yet it remains inadequately represented in global models. In this study, an efficient and integrated numerical model, the Moving-Grid Permafrost Model (MVPM) was used to simulate the permafrost thermal regime in West Kunlun (WKL), which is approximately 55,669 km² in northwest Qinghai-Tibet Plateau with extreme arid climate conditions. We employed clustering approaches and parallel computing techniques to enhance computational efficiency. The model forcing data, remote-sensing-based land surface temperature (LST) dating back to 1980 with a spatial resolution of 1 km×1 km and a temporal resolution of 1month, was constructed using machine learning techniques that integrate field observations, satellite data and reanalysis products. Our simulations achieved high accuracies of ±0.25 °C for ground temperature and ±0.25 m for active layer thickness, significantly outperforming previous simulations reported to date. The results indicated that the WKL experienced a pronounced warming trend in LST, with an average increase of 0.40 °C per decade from 1980 to 2022. The responses of the permafrost regime to climate warming were closely related to the original thermal conditions shaped by historical climatic evolution. These responses exhibited a distinct altitude-dependent spatial variation and differed according to soil stratigraphic types. Despite the thermal warming trend, the areal extent of permafrost remained relatively stable across the WKL region over the past 43 years, reflecting the slow and lagged response of permafrost to climate warming. These findings are essential for enhancing our understanding of permafrost thaw trajectories, and improving projections of potential future consequences of permafrost degradation with greater accuracy.