Artificial intelligence is deeply affecting the technology of the future and the present. The most talked-about name in the world of artificial intelligence recently has been the Chinese company DeepSeek. So, what is DeepSeek? How is DeepSeek used?
WHAT IS DEEPSEEK?
DeepSeek is a China-based artificial intelligence research laboratory focused on developing open-source large language models (LLMs). The company was founded in 2023 by Liang Wenfeng, the founder and manager of the hedge fund High-Flyer, to advance its work in the field of artificial intelligence. Previously, High-Flyer was known for developing AI-based trading algorithms.
DeepSeek Coder (November 2023): Specifically targeting coding tasks, this model was offered for free to both researchers and commercial users. The model was released as open-source under the MIT license.
MODELS DEVELOPED
DeepSeek LLM (November 2023): With 67 billion parameters, this model was designed to compete with large language models like GPT-4. However, it faced some challenges in terms of computational efficiency and scalability. The chatbot version of the model, DeepSeek Chat, was also released during the same period.
DeepSeek-V2 (May 2024): Released at a lower cost compared to its competitors (2 RMB per million output tokens), this model ranked seventh in the University of Waterloo Tiger Lab rankings.
DeepSeek-V3 (December 2024): With 671 billion parameters, this model was trained for approximately 55 days on a massive dataset of 14.8 trillion tokens. The total cost of the training process was 5.58 million US dollars. With its performance, the model surpassed Llama 3.1 and Qwen 2.5, reaching a level equivalent to GPT-4o and Claude 3.5 Sonnet.
DeepSeek R1-Lite-Preview (November 2024): This model specializes in logical inference, mathematical reasoning, and real-time problem solving. It has demonstrated performance similar to OpenAI's o1 model.
News Source: 12punto
