The integration of Emotional Intelligence into Large Language Models is an innovative advance in the search for human-like and sensitive interaction. This review synthesizes insights from a diverse peer-reviewed research articles that delve into methodologies, benchmarks, datasets, and fine-tuning techniques in furthering the emotional understanding of LLMs, its affective recognition, and empathetic responsiveness. In particular, emotionally intelligent LLMs have proved helpful in mental health support tasks by offering empathetic, context-sensitive counseling that increases the engagement of users and positive outcomes. The present review tries to, answer certain key research questions related to domain-specific applications, evaluation metrics, and technical challenges, and give a general overview of the current progress within this field and identify future research opportunities. It also highlights some lingering challenges: cultural sensitivity, multimodal integration, and real-time emotional adaptation.