This article presents the design and evaluation of a Rectangular Dielectric Resonator Antenna (RDRA) for sub-THz applications, specifically at a frequency of 0.49 THz. The antenna exhibits promising performance with a return loss of −35.45 dB, a bandwidth of 351 GHz, and an efficiency of approximately 85%. With a compact size of 59 μm × 59 μm, the RDRA offers a high-performance solution for next-generation wireless communication systems. The antenna's performance was assessed through software simulations, an RLC equivalent circuit model, and Machine Learning (ML) techniques. The resonance frequencies predicted by the Resistor, Inductor, Capacitor (RLC) model, simulated using Advanced Design System (ADS) Agilent software, align closely with those from other modeling tools, validating the accuracy of the design. Furthermore, five regression-based ML models were developed to predict the dielectric antenna's gain, with the XGB regression model demonstrating the best prediction accuracy. The study also explores the efficacy of the ML models using several evaluation metrics, including variance score, R-squared, Mean Square Error (MSE), and Root Mean Square Error (RMSE). The results highlight the potential of the RDRA for sub-THz applications, such as ultra-fast wireless communication and high-resolution imaging, with ML enhancing the design optimization process.