This study investigates how Large Language Models (LLMs) can be used to deliver formative feedback to first-year students in an introductory programming course and to provide instructors with insights into overall class performance. It aims to evaluate the quality of feedback generated by these models and compare their performance with human evaluators on programming assignments. In addition, it explores the user experience of interacting with the AI tutor. Findings underscore the potential of advanced AI to enhance personalized learning and formative assessment in educational settings.