A new study published on arXiv evaluates the performance of five leading generative AI systems, including ChatGPT 5.2, DeepSeek-V3, Gemini 2.5 Flash, Claude Sonnet 4.5, and M365 Copilot, on introductory object-oriented programming assessments. The research found that all evaluated AI models significantly outperformed the average student cohort, frequently achieving perfect scores on complex coding tasks. However, the models still exhibited limitations, occasionally producing non-compiling code and struggling with advanced object-oriented concepts like abstract classes and certain inheritance scenarios, as well as graphics-related questions. The study also noted improvements in AI performance compared to the previous year, while identifying persistent error patterns. AI
IMPACT Indicates GenAI's growing capabilities in coding education and assessment design, while highlighting areas needing further development.
RANK_REASON Academic paper evaluating AI performance on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]
- ChatGPT 5.2
- Claude Sonnet 4.5
- DeepSeek-V3
- Gemini 2.5 Flash
- generative artificial intelligence
- M365 Copilot
- object-oriented programming
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
