Beijing, April 24 (es.investing.com) - The Chinese company DeepSeek announced on Friday the preliminary and open-source launch of its new artificial intelligence model, DeepSeek V4, highlighting that its performance is comparable to leading U.S. models in areas such as reasoning, general knowledge, and capabilities for intelligent agents. According to a statement released on its official WeChat account, DeepSeek has divided its system into two versions: V4-Pro and V4-Flash. The company claims that the version focused on advanced reasoning ranks among the global leaders. DeepSeek had already generated significant anticipation in early 2025 with its R1 model, which managed to compete with U.S. alternatives at a significantly lower cost. According to the technical data published, the V4-Pro-Max outperforms models like Claude-Opus-4.6-Max (Anthropic) and GPT-5.4-xHigh (OpenAI) in knowledge tests and is on par in software engineering tasks. In terms of reasoning, the company states that it is only slightly behind Google’s Gemini-3.1-Pro-High. One of the main innovations of V4 is its context window of up to one million tokens, thanks to a renewed attention architecture that, according to DeepSeek, reduces computational and memory resource consumption compared to traditional methods. The model has been optimized to work with agent frameworks such as Claude Code, OpenClaw, OpenCode, and CodeBuddy, showing improvements in coding tasks and document generation. DeepSeek has not revealed details about the hardware used to train the V4. It is worth noting that the R1 model was trained in just 55 days with a budget of $5.57 million, a figure much lower than that of its U.S. competitors, reigniting the debate about the feasibility of developing advanced AI with fewer resources. In recent months, the company has expanded its portfolio with new versions of R1 and systems like DeepSeek-OCR and V3.2, the latter also presented as comparable to OpenAI and Google models in certain aspects. The launch of V4 comes amid strong competition in the Chinese AI sector, where giants like Baidu, Alibaba, Tencent, ByteDance, and MiniMax have accelerated the development of models and agents, in the context of technological rivalry with the U.S., restrictions on access to advanced semiconductors, and Beijing's strategy to strengthen technological self-sufficiency.