DeepSeek vs ChatGPT: Prospects and Challenges

Inhye Jin¹Jonathan Alan Tangsrivimol^2,3Erfan Darzi²Hafeez Ul Hassan Virk⁴Zhen Wang^5,6Jan Egger⁷Sean Hacking⁸Benjamin S Glicksberg⁹Markus Strauss^10,11*Chayakrit Krittanawong, MD, FACC¹²

• ¹College of Medicine, Yeungnam University, Gyeongsan, North Gyeongsang, Republic of Korea
• ²Chulabhorn Royal Academy, Bangkok, Thailand
• ³New York Presbyterian Hospital, New York, New York, United States
• ⁴Computer Science and Artificial Intelligence Laboratory, Schwarzman College of Computing, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States
• ⁵Case Western Reserve University, Cleveland, Ohio, United States
• ⁶Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Jacksonville, Florida, United States
• ⁷Essen University Hospital, Essen, North Rhine-Westphalia, Germany
• ⁸Department of Pathology, Grossman School of Medicine, New York University, New York, New York, United States
• ⁹Icahn School of Medicine at Mount Sinai, New York, New York, United States
• ¹⁰University Muenster, Münster, Germany
• ¹¹Faculty of Health, Witten/Herdecke University, Witten-Herdecke, North Rhine-Westphalia, Germany
• ¹²Langone Medical Center, New York University, New York City, New York, United States

DeepSeek has introduced its recent model DeepSeek-R1, showing divergence from OpenAI's ChatGPT, suggesDng an open-source alternaDve to users. This paper analyzes the architecture of DeepSeek-R1, mainly adopDng rule-based reinforcement learning (RL) without preliminary supervised fine-tuning (SFT), which has shown beNer efficiency. By integraDng mulD-stage training along with cold-start data usage before RL, the model can achieve meaningful performance in reasoning tasks along with reward modeling opDmizing training process. DeepSeek shows its strength in technical, reasoning tasks, able to show its decision-making process through open source whereas ChatGPT shows its strength on general tasks and areas requiring creaDveness. Despite the groundbreaking developments of both models, there is room for improvement in AI landscape and maNers to be handled such as quality of data, black box problems, privacy management, and job displacement. This paper suggests the future of AI, expecDng beNer performance in mulD-modal tasks, enhancing its effecDveness in handling larger data sets, enabling users with improved AI landscapes and uDlity.