Agricultural price forecasting is a crucial area of research, particularly in regions like Delhi, India, where volatile prices of essential crops such as tomato, onion, and potato (TOP) can significantly impact food security and economic stability. This study undertakes a comprehensive analysis of eleven advanced time series forecasting models, including long short-term memory (LSTM), recurrent neural network (RNN), gated recurrent unit (GRU), temporal convolutional network (TCN), CNN-LSTM hybrid model, production, estimation, and crop assessment division (PECAD), bidirectional LSTM (BiLSTM), bidirectional GRU (BiGRU), stacked LSTM, and the attention-based convolutional neural network with optimized bidirectional long short-term memory (ACNN-OBDLSTM). The objective is to determine the most effective model for predicting TOP crop prices in the Azadpur Market, Delhi, India. The LSTM demonstrates superior precision in forecasting TOP crop prices, with its low RMSE, MAE, MAPE, and high R-squared scores making them a promising tool for forecasting crop prices in the Delhi region. The LSTM displayed the best MAPE score of 4.05% for tomato crops, 3.9% for onion crops, and 1.64% for potato crops.