Chinese artificial intelligence company DeepSeek has introduced its latest flagship AI system, marking a major upgrade roughly one year after its earlier breakthrough that drew global attention.
The new release includes a family of models known as DeepSeek V4, featuring versions designed for both high-performance tasks and lighter, faster use. The company says the updated system delivers major improvements in reasoning ability, coding performance, and multi-step task execution, allowing the AI to handle more complex instructions with reduced human input.
A key feature of the new model is its improved “hybrid attention” design, which enhances the system’s ability to process and retain long sequences of information. This allows it to work more effectively with large documents and extended conversations compared to previous versions.
DeepSeek also emphasized that the new AI is built for broad hardware compatibility, supporting both leading U.S. chips and China’s domestic processors. This reflects growing efforts in the country to strengthen independence in advanced computing technology.
Early reports suggest the flagship model is competitive with top global AI systems, particularly in programming and analytical reasoning tasks, while also being more cost-efficient to operate. Alongside the main model, the company has also released a lighter version aimed at faster, lower-cost applications.
The launch highlights the accelerating global race in artificial intelligence, as major tech firms continue to compete for leadership in next-generation AI capabilities.