With the rapid advancement of AI-driven language models, their role in STEM and computing education continues to evolve. Following our previous study on ChatGPT-3.5’s performance in assisting students with introductory web programming tasks, this paper extends the analysis by comparing other available alternatives, including DeepSeek and Claude. Our focus remains on evaluating these models as teaching assistants and code generators for solving standard exam-level programming problems.

We assess each model’s ability to generate correct and complete solutions for fundamental web development tasks involving HTML, CSS, JavaScript, and elementary PHP. The evaluation follows the same methodology as our previous research—tasks are presented as exam-style prompts, with AI-generated solutions graded using the same criteria applied to student submissions. We analyze the initial responses, improvements after refinement, and the models’ adaptability to feedback.

The results demonstrate that most current AI language models perform competently in generating functional web programming code, with varying strengths across different programming languages and concepts. Building upon our previous findings with ChatGPT-3.5, we observe that several AI alternatives now offer comparable assistance capabilities, including code generation, for introductory web programming. This research provides insights into the growing integration of AI assistants in educational contexts and development environments, while discussing implications for pedagogical approaches and academic assessment in computing education.

Keywords: ChatGPT, Web programming, JavaScript, HTML, CSS, PHP, Gemini, DeepSeek, Claude, Cursor.