This review explores the use of large language models (LLMs) as virtual tutors in education. LLMs, based on transformer architectures, have gained significant attention for their ability to generate and evaluate educational materials. The review focuses on systems specifically designed for educational purposes, involving students and teachers in their design or experimental plans. It highlights the most common applications of LLMs, such as automatic question generation, answer grading, and code explanation. The most popular models include GPT-3, BERT, and T5. The review also discusses the challenges and ethical considerations of using LLMs in education, including issues of academic integrity and the need for transparency. The study identifies 29 relevant works that meet the criteria, with a focus on reproducible and ethically sound applications. The review concludes that while LLMs offer promising opportunities for enhancing educational practices, further research is needed to address the challenges and ensure their effective integration into educational curricula. The review also notes the potential of LLMs in various educational tasks, including text summarization, learning software, and problem resolution. The study emphasizes the importance of collaboration between researchers, developers, and end-users to advance the use of LLMs in education.This review explores the use of large language models (LLMs) as virtual tutors in education. LLMs, based on transformer architectures, have gained significant attention for their ability to generate and evaluate educational materials. The review focuses on systems specifically designed for educational purposes, involving students and teachers in their design or experimental plans. It highlights the most common applications of LLMs, such as automatic question generation, answer grading, and code explanation. The most popular models include GPT-3, BERT, and T5. The review also discusses the challenges and ethical considerations of using LLMs in education, including issues of academic integrity and the need for transparency. The study identifies 29 relevant works that meet the criteria, with a focus on reproducible and ethically sound applications. The review concludes that while LLMs offer promising opportunities for enhancing educational practices, further research is needed to address the challenges and ensure their effective integration into educational curricula. The review also notes the potential of LLMs in various educational tasks, including text summarization, learning software, and problem resolution. The study emphasizes the importance of collaboration between researchers, developers, and end-users to advance the use of LLMs in education.