DeepSeek is a Chinese artificial intelligence (AI) company specializing in the development of open-source large language models (LLMs). Established in July 2023 by Liang Wenfeng, who also co-founded the hedge fund High-Flyer, DeepSeek operates out of Hangzhou, Zhejiang, China. The company is privately held and employs fewer than 200 individuals, many of whom are recent graduates from top Chinese universities.
Key Developments and Releases:
DeepSeek Coder (November 2023): An open-source model tailored for coding tasks, marking the company’s initial foray into AI model development.
DeepSeek LLM (November 2023): A 67-billion-parameter model designed to compete with other large language models, enhancing the company’s presence in the AI community.
DeepSeek-V2 (May 2024): Notable for its strong performance and cost-effectiveness, this release intensified competition in the Chinese AI market, prompting major tech firms to adjust their pricing strategies.
DeepSeek-Coder-V2: An advanced model with 236 billion parameters and a context length of up to 128,000 tokens, aimed at addressing complex coding challenges.
DeepSeek-V3 (December 2024): A 671-billion-parameter model demonstrating impressive performance across various benchmarks while utilizing fewer resources compared to its peers.
DeepSeek-R1 (January 2025): Focused on reasoning tasks, this model challenges existing AI systems with its advanced capabilities.
On January 20, 2025, DeepSeek released its first free chatbot app for iOS and Android, based on the DeepSeek-R1 model. Within a week, it surpassed OpenAI’s ChatGPT to become the top-rated free application on Apple’s App Store in the United States. This rapid ascent has raised questions about American dominance in AI technology.
DeepSeek’s models are developed amid U.S. sanctions aimed at restricting China’s access to advanced semiconductor chips. Despite these challenges, the company has managed to create AI systems that are both efficient and cost-effective. For instance, the DeepSeek-R1 model was developed at a cost of approximately $6 million, significantly lower than the $100 million reportedly spent on OpenAI’s GPT-4 in 2023.
The company adopts an open-source approach, making its AI algorithms, models, and training details freely available for use and modification. However, reports indicate that DeepSeek’s models apply content restrictions in line with local regulations, limiting responses on topics deemed politically sensitive by the Chinese government.
DeepSeek’s emergence has significant implications for the global AI landscape, challenging established players and prompting discussions about innovation, competition, and geopolitical dynamics in the technology sector.
