ChinaŌĆÖs AI upstart DeepSeek drops new model. Will it make waves like last year?

DeepSeek, a rising star in China's AI sector, has introduced a preview of its eagerly awaited model, V4, positioning it to challenge established competitors like OpenAI, Anthropic, and Google. This announcement comes just a year after the company, based in Hangzhou, made headlines with its previous model, R1, which surprised the global AI landscape with its impressive performance and affordability. The new V4 model boasts significant enhancements in reasoning and agentic capabilities, enabling it to perform tasks autonomously, such as coding. Additionally, the model is designed to efficiently handle larger volumes of tokens, which are essential for AI systems to interpret instructions effectively. DeepSeek's rapid rise in the AI domain began with its groundbreaking R1 model in 2025, which not only showcased near top-tier performance but also did so at a fraction of the cost of its American counterparts. This achievement had a ripple effect, causing a decline in American AI stocks and igniting discussions about the substantial investments in data center infrastructure. Moreover, it heightened confidence in China's technological advancements, intensifying the rivalry with the U.S. However, analysts predict that V4 may not generate the same market excitement as R1 did. Ivan Su, a senior equity analyst at MorningStar, noted that while R1 was a shocking revelation, V4 is viewed as a continuation of that trend, and the market has already adapted to the competitive pricing and capabilities of Chinese AI solutions, which may limit investor reactions this time. In line with its predecessors, V4 remains open source, making it accessible to a wide audience, unlike many American models. This strategy is integral to China's approach to competing with the U.S., as it allows for rapid adoption and deployment across various sectors, from e-commerce to robotics. The necessity for such strategies is compounded by the limited access to advanced chips due to U.S. export restrictions, which have compelled Chinese developers to collaborate with domestic chip manufacturers. To support the computational demands of V4, DeepSeek has teamed up with Huawei, leveraging its Supernode technology and Ascend 950 chips to boost processing power. Wei Sun, a principal analyst at Counterpoint Research, emphasized that V4's use of domestic chips could lead to greater domestic adoption and faster global AI development compared to R1, which relied on Nvidia hardware. While top American AI models like Claude, ChatGPT, and Gemini currently dominate the market, Chinese firms are making significant strides in the open-source arena. DeepSeek claims that V4 exhibits superior agentic coding abilities among open models and achieves exceptional reasoning capabilities, though it still acknowledges a gap compared to leading industry models like Gemini. Despite its achievements, DeepSeek faces scrutiny, with accusations from competitors like Anthropic and OpenAI alleging that the startup has unlawfully extracted capabilities from their models. Furthermore, U.S. officials have raised concerns about foreign entities, particularly from China, engaging in large-scale efforts to replicate advanced AI models from American companies. While these accusations have not specifically named DeepSeek, they highlight the ongoing tensions between the two global superpowers. CNN has reached out to DeepSeek for a response to these allegations.