According to sources familiar with the matter, Hangzhou AI startup DeepSeek is accelerating the development of an AI model with advanced agent capabilities, targeting US tech giants like OpenAI. This model, centered around powerful autonomous execution capabilities, can complete complex, multi-step operations with minimal user input and self-optimize based on historical experience. Founder Liang Wenfeng is personally overseeing its development, with a planned release in the fourth quarter of 2025, demonstrating the ambitions of Chinese companies in the field of AI agent technology.
DeepSeek, incubated by quantitative giant Huanfang Quantitative, boasts a team with extensive expertise in AI. Its previously released open-source model, DeepSeek-V3, has demonstrated outstanding performance in Chinese language tasks. This newly developed agent technology is considered a key milestone in AI development and can be widely applied in areas such as agricultural decision-making, content generation, and intelligent customer service, significantly reducing human intervention. Industry analysts indicate that AI agents are driving the transformation of enterprise software from "feature stacking" to "intelligent collaboration," with the Chinese market projected to reach 852 billion yuan by 2028, with an average annual growth rate of 72.7%.
Facing competition from OpenAI, DeepSeek differentiates itself with its open-source strategy and Chinese-specific capabilities. If its new model is released as scheduled, it may seize the initiative in the global AI agent competition and further rewrite the technological landscape.
