Agriculture is the most significant sector that demands strategic attention from every nation aspiring to achieve developed status. The most transformative advancement in this sector comes from integrating large language models (LLMs) to optimize agricultural practices such as soil formation, irrigation, harvesting, and crop rotation. To address the challenges of the sector, such as prediction of crop diseases, availability of resources, price fluctuations, soil degradation, and inadequate market infrastructure, several data-driven strategies are adopted frequently. This chapter initially introduces the foundational architecture of LLMs, including transformers, self-attention mechanism, and positional encoding. Further, the functioning of LLMs is demonstrated by analyzing agricultural context sentences with a pretrained BERT model, illustrating their practical applications. Several special agricultural LLMs, viz. AgriBERT and FarmBERT, which are designed to address the challenges of advanced agricultural text processing tasks, are discussed in detail. Finally, the chapter discusses the solutions and challenges of data availability and domain specificity faced during the execution of LLMs for agricultural data. The chapter concludes by summarizing future pathways for research and development in the area of agricultural LLMs to drive innovations in the agriculture sector.