DeepSeek is reportedly preparing to launch a next-generation artificial intelligence (AI) agent by the end of 2025, designed to complete complex tasks with little user guidance and capable of learning from previous actions to improve its performance.
the Chinese AI company first attracted global attention earlier this year with the launch of the R1 model, which featured advanced reasoning capabilities at a development cost of only around US$6 million.
This figure is significantly lower than that of large models from OpenAI or Google. Another advantage is that R1 is designed as an open-source model, making its source code freely accessible to developers. R1's arrival has shaken up the "bigger is better" trend that has prevailed among several tech giants in the United States.
Since then, DeepSeek has adopted a cautious approach. The company's founder, Liang Wenfeng, temporarily delayed the launch of the R2 model to allow for technical refinements and to align his activities at High-Flyer Asset Management. This strategy differs from Alibaba and Tencent, which accelerated the release of their AI models.
