Option pricing has historically been a significant challenge in financial markets and recent progress in deep learning methodologies has intensified interest in creating precise and consistent models for pricing and prediction. However, the accuracy and reliability of these models depend significantly on effective hyperparameter tuning. We carried out an in depth examination of diverse deep learning models to assess their predictive performance in option pricing. We specifically examine their behavior when optimized using Bayesian optimization (BO) and compare the results with models that do not incorporate BO. We analyze S&P 500 option price data to evaluate deep learning models’ performance in option pricing and the impact of hyperparameter tuning on predictive accuracy. To rigorously evaluate model performance, we employ multiple error metrics and conduct a paired t-test to evaluated the significance of improvements achieved through BO. Our findings demonstrate that Bayesian optimization significantly enhances model accuracy, leading to more precise option pricing predictions. The results underscore the importance of systematic hyperparameter tuning in deep learning (DL) applications for financial modeling. Finally, the article concludes by highlighting the potential scope for future work.