Chinese AI startup DeepSeek has announced plans to release five new code repositories next week, reinforcing its commitment to open-source artificial intelligence.
The company, which gained global attention with its R1 reasoning model rivaling Western AI systems, described the release as ‘small but sincere progress’ towards full transparency. These repositories, tested in real-world applications, will provide essential infrastructure to support the AI models DeepSeek has already made public.
DeepSeek has set itself apart in China‘s AI sector by embracing open-source practices, a rare move in a market that typically favours closed-source models.
Founder Liang Wenfeng has emphasised the cultural significance of open-source over commercial gains, highlighting the satisfaction that comes from driving innovation and earning industry respect.
His approach has sparked global interest, particularly after the release of the R1 model, known for its strong performance and cost efficiency.
The company has also recently introduced a new algorithm, Native Sparse Attention (NSA), designed to improve efficiency in long-context training and inference.
DeepSeek’s popularity has surged, becoming China’s leading chatbot service with 22.2 million daily active users, surpassing long-established platforms like Douban. The growing user base and commitment to open-source are positioning DeepSeek as a major player in the global AI landscape.
For more information on these topics, visit diplomacy.edu.