<p>The next generation of optical neuromorphic systems demands devices with non-volatile programmability, ultra-compact footprints, and exceptionally low static energy consumption—challenges that silicon-based technologies struggle to overcome. In this study, we present an innovative solution by integrating phase-change materials (PCM) with plasmonic nanoscale structures, effectively addressing these limitations. The proposed device features a nanoscale filter based on a hybrid X-shaped and square resonator (XSR) coupled to orthogonal horizontal and vertical waveguides; one waveguide transmits the input signal, while the other modulates synaptic weight, with scalability to all synapses of a neuron. This intelligent design minimizes device size and significantly reduces crosstalk between waveguides. Structural symmetry enables adaptive waveguide utilization tailored to application-specific needs. The unique combination of cross and square geometries ensures optimal resonance wavelength alignment with the PCM’s peak performance range without increasing footprint. The device achieves an ultra-compact footprint of 0.14 µm² and a remarkable optical contrast of 74%, delivering efficient and stable synaptic functionality. Moreover, reducing the PCM volume results in an ultra-low static energy consumption of 8 pJ, marking a significant advancement over existing counterparts. These findings open new horizons for developing scalable, energy-efficient optical neural networks that closely emulate the architecture and function of the human brain.</p>