Efficiency and safety of the DeepSeek R1 model compared to OpenAI models

Mikita Piastou

Аннотация


The development of advanced Large Language Models (LLMs) has led to a growing importance in the selection of a particular model for use in real-world scenarios, including science, education, and information handling. Consequently, a comparative study of the new DeepSeek R1 model and OpenAI models is relevant. The objective of this paper is to evaluate the performance and security of the DeepSeek R1 model against recent OpenAI models in terms of prediction accuracy, resilience to attacks, and ethics and privacy compliance. The findings demonstrate that DeepSeek R1 can perform better in tasks requiring detailed and sound logical reasoning, outperforming o1 models, but is inferior to o3 models, especially in terms of mathematical ability. The quality of DeepSeek R1 application in various scientific and practical fields is a subject that requires further research.

Полный текст:

PDF

Ссылки

  • На текущий момент ссылки отсутствуют.


ISSN 2414-4487