CHATBOT

Large Language Models (LLMs) have started a cyber war of achieving Artificial General Intelligence (AGI).  ChatGPT and Deep Seek R1 are prominent large language models that have revolutionized the tech world by their potential, accuracy, achievements and efficiency. These models are reshaping the tech industry. In this article we’ll compare both the models at various proficiency levels. Developmental Origin: ChatGPT was launched in November 2022, by American research organization known for its advancements in artificial intelligence named OpenAI. OpenAI has several other models such as GPT-4, GPT-4o mini, o1 and o1 mini, GPT-4 turbo, GPT-3.5 Turbo, Whisper, Dall-e and much more. R1 was launched in January 2025, developed by Deep Seek a Chinese Company located in Hangzhou. Deep Seek is an AI company developing models to achieve Artificial General Intelligence (AGI). It has several other models like; DeepSeek-V3, DeepSeek-V2.5, DeepSeek-R1-lite, etc. Performance and Efficiency:   The above image demonstrates the Accuracy/Percentile of models; DeepSeek-R1, OpenAI-o1, DeepSeek-R1-32B, OpenAI-o1-mini, and DeepSeek-V3. The first parameter taken in the above image AIME represents accuracy in solving advanced math problems. The analysis shows a 79.8 % score of DeepSeek-R1 while OpenAI-o1 model 79.2%. The Codeforces percentile represents programming and problem-solving capabilities. DeepSeek-R1 achieved an accuracy of 96.3% while OpenAI-o1 scored 96.6%. GPQA Diamond percentile demonstrating Question-Answer tasks. Here OpenAI-o1 achieved 75.7% surpassing DeepSeek-R1 which achieved 71.5%. The MATH-500 benchmarks math-solving capabilities. DeepSeek-R1 is ahead scoring a percentile of 97.3%, while OpenAI-o1 scored 96.4%. MMLU evaluates understanding across multiple knowledge domains. OpenAI-o1 outperforms by achieving 91.8% while DeepSeek-R1 scored 90.8%. SWE-bench Verified tests software engineering-related challenges. 49.2% scored by DeepSeek-R1 while 48.9% scored by OpenAI-o1. DeepSeek-R1 consistently outperforms major benchmarks as compared to other models. OpenAI models; o1 and o1-mini also performed well but lag slightly as compared to DeepSeek-R1. Cost Efficiency: DeepSeek-R1 excelled in cost efficiency as being approximately 27 times cheaper per token compared to OpenAI’s models. Deep Seek price for one million tokens starts at $0.14, which is much lower than OpenAI’s model, which costs $7.50 for the same token volume. DeepSeek-R1 model was trained at 6 million dollars, while GPT-4 cost over 100 million dollars, and Gemini reportedly cost over 200 million dollars.  Openness and Transparency: DeepSeek-R1 is an open-source model, providing insight into their algorithms, architecture, and training processes. This allows developers, researchers, and organizations to access, modify, and build upon the model without restrictions. ChatGPT is a proprietary model, its underlying code, training methodologies, and data are not publicly accessible. While it provides APIs and user-friendly interfaces for implementation, users cannot directly access or alter the model’s inner workings. Censorship and Criticism: DeepSeek-R1 has faced criticism regarding censorship and data privacy. The model reportedly employs censorship mechanisms for topics considered politically sensitive in China, such as the 1989 Tiananmen Square protests and the status of Taiwan. ChatGPT has also faced scrutiny over data privacy and content moderation, it operates under different regulatory frameworks and has not been associated with the same level of censorship concerns. Conclusion Scraping Solution ’s research and some lab work concludes that both the models have a significance in the advancements in Artificial Intelligence. DeepSeek-R1 has offered a cost-effective, open-source alternative with comparable performance while ChatGPT benefits from the extensive resources and research backing of OpenAI.