<p>Thermophilic proteins maintain their structure and function at high temperatures, making them widely useful in industrial applications. Due to the complexity of experimental measurements, predicting the melting temperature (<i>T</i>_m) of proteins has become a research hotspot. Previous methods rely on amino acid composition, physicochemical properties of proteins, and the optimal growth temperature (OGT) of hosts for <i>T</i>_m prediction. However, their performance in predicting <i>T</i>_m values for thermophilic proteins (<i>T</i>_m&gt;60 °C) are generally unsatisfactory due to data scarcity. Herein, we introduce TmPred, a <i>T</i>_m prediction model for thermophilic proteins, that combines protein language model, graph convolutional network and Graphormer module. For performance evaluation, TmPred achieves a root mean square error (RMSE) of 5.48 °C, a pearson correlation coefficient (<i>P</i>) of 0.784, and a coefficient of determination (<i>R</i>^<i>2</i>) of 0.613, representing improvements of 19%, 15%, and 32%, respectively, compared to the state-of-the-art predictive models like DeepTM. Furthermore, TmPred demonstrated strong generalization capability on independent blind test datasets. Overall, TmPred provides an effective tool for the mining and modification of thermophilic proteins by leveraging deep learning.</p>