Pricing options accurately is a fundamental challenge in quantitative finance. While traditional models like Black–Scholes have been widely used, their reliance on restrictive assumptions limits their applicability. Similarly, modern deep learning techniques, though powerful, often struggle to reconcile physical consistency with the flexibility needed to adapt to real-world data. This study introduces a hybrid framework that combines physics-informed neural networks (PINNs) and Fourier neural operators (FNOs) to overcome these challenges. The hybrid model exploits the strengths of PINNs for embedding the Black–Scholes equations and FNOs for capturing complex global relationships, achieving unprecedented accuracy and robustness. Validated on synthetic datasets of local volatility surfaces, the hybrid approach demonstrates up to a 50% reduction in Mean Squared Error (MSE) compared to stand-alone models. Visual analyses further highlight its ability to combine mathematical rigour with practical adaptability, making it well suited to real-world applications.