Chinese startup DeepSeek on Friday unveiled a preliminary version of its new artificial intelligence model, V4, continuing the trajectory that brought the company onto the global stage a year ago. In 2025, the then little-known Hangzhou-based firm made a splash with its R1 model, which could compete with leading Western platforms at a significantly lower cost. According to Freddy Miller, Senior Analyst at NEWSCENTRAL, the success of R1 sent a signal to the global industry and proved that Chinese AI can reach the level of world leaders.
The new V4 model features enhanced logical reasoning and agent-based programming capabilities, allowing it to perform tasks on behalf of users, including coding and managing autonomous processes. The model is also optimized to handle large token volumes, which is critical for large-scale computational tasks. These capabilities make V4 a promising tool for companies looking to integrate AI into real business processes.
A key distinction of V4 is its use of domestic chips from Huawei and Cambricon, providing the necessary computing power without reliance on American processors. Lucas Grant, an analyst in the semiconductor industry and manufacturing strategies at NEWSCENTRAL, explains that the combination of Ascend 950 and Cambricon chips allows DeepSeek to operate large language models on local infrastructure, bypass export restrictions, and accelerate AI deployment in industrial and corporate applications.
V4 maintains an open-source strategy, which accelerates the dissemination and integration of the model into research and commercial projects. Open models create conditions for building a global AI ecosystem and allow companies to quickly adapt the technology to their needs.
At the same time, questions remain regarding the legality of technology use. Companies like Anthropic and OpenAI have previously expressed concerns about the potential extraction of capabilities from their models, and White House representatives have pointed to campaigns by foreign organizations to reverse-engineer advanced AI algorithms. This underscores the growing competition between the U.S. and China and increases the risks associated with international collaboration in technology.
Technically, V4 surpasses other open models in world knowledge and agent-based programming ability, although it lags behind industry leaders such as Google’s Gemini. Experts see this as an opportunity for China to strengthen its position in the open-model segment and accelerate AI deployment in industry and services, even though global leadership still belongs to Western proprietary systems.
From a strategic perspective, V4 demonstrates the maturity of China’s AI ecosystem. The combination of openness, domestic chips, and autonomous features reduces costs and speeds up solution deployment. According to Freddy Miller, in the next 2–3 years, V4 will become a key driver of AI adoption in Asian countries and gradually increase the influence of Chinese technologies on the global market.
NEWS CENTRAL notes that for businesses and investors, the key takeaways are high performance at moderate costs, independence from foreign processors, and broad integration capabilities. It is recommended to monitor the development of V4, assess the potential of strategic partnerships with Chinese startups, and leverage open platforms to accelerate AI technology adoption in the real sector.