A study evaluated two contemporary large language models (LLMs), ChatGPT (GPT-5.2) and Gemini (Ultra/Pro), in radiation oncology between 1 January and 30 January 2026. The research aimed to compare their performance and explore the strengths and limitations of benchmark-based evaluation approaches, including challenges related to clinical interpretation, reference-answer validity, and assessment of open-ended responses. The study noted that while earlier-generation models like ChatGPT-3.5 could answer a substantial percentage of radiation oncology-related questions with notable limitations, LLM technology has advanced considerably, with newer generations showing higher performance. However, the integration of LLMs into routine clinical practice remains limited by challenges such as accuracy, bias, and the need for careful validation. The authors emphasized that their application in oncology requires careful validation, expert oversight, and domain-specific adaptation to ensure safe and effective use.