Accurate prediction of foreign exchange rates is crucial for effective decision-making in global financial markets. This study investigates the performance of several advanced deep learning models—namely, convolutional neural networks, long short-term memory networks, gated recurrent units and their hybrid combinations—in forecasting foreign exchange rates. The models were developed and trained on historical Forex data sourced from Yahoo Finance and Investing.com, covering daily, weekly and monthly intervals. Each model’s performance was rigorously evaluated using multiple metrics, including mean absolute error, root mean squared error and R-squared to ensure a complete assessment. The analysis included comparisons of standalone and hybrid models to determine their effectiveness in capturing market trends and patterns. Results demonstrated that hybrid models showed superior predictive accuracy across different timeframes. These findings contribute to improving Forex prediction techniques and provide valuable insights for optimising trading strategies, leading to more informed decision-making.