Chinese artificial intelligence company DeepSeek has introduced a new AI model, further intensifying competition with leading U.S. technology firms such as OpenAI and Google. The launch comes after the company’s earlier breakthrough model in 2025 reportedly caused significant disruption in global AI markets.

DeepSeek first gained international attention in January 2025 when it released its AI chatbot “R1,” which quickly attracted global users and drew comparisons with OpenAI’s ChatGPT. At the time, the company claimed that the model was developed at a remarkably low cost of under $6 million, while still delivering performance comparable to leading Western AI systems. This announcement reportedly unsettled parts of the U.S. tech industry and contributed to volatility in AI-related markets.

Now, more than a year later, DeepSeek has launched its next-generation models: DeepSeek V4 Pro and DeepSeek V4 Flash. A preview version was released on April 24, and the company claims the new systems are competitive with advanced AI models developed by OpenAI and Google.

Like its previous versions, the new models are based on an open-source framework, allowing developers to freely access, modify, and integrate the system into their own applications. This approach is seen as one of DeepSeek’s key strategies for rapid adoption and innovation.

According to performance benchmarks shared by the company, the V4 Pro model has shown strong results in mathematics and coding tasks, reportedly outperforming several other open models in these categories. In general knowledge tests, it is said to trail only Google’s Gemini 3.1 Pro, which itself is a closed-source model.

DeepSeek also stated that its latest model performs only slightly behind OpenAI’s GPT-5.4 and Google’s Gemini 3.1 Pro in overall capability, estimating a technological gap of roughly three to six months compared to leading U.S. systems. The V4 Flash version offers similar capabilities but is optimized for faster responses and lower operational cost.

The company behind these developments was founded in May 2023 by Liang Wenfeng in China’s Guangdong province. Unlike many commercial AI firms, DeepSeek was designed primarily as a research-driven organization rather than a profit-focused enterprise.

Liang Wenfeng has previously emphasized that the company’s goal is not financial gain but technological advancement. He has described DeepSeek as an attempt to push forward the entire AI ecosystem rather than compete purely on commercial grounds.

Funding for DeepSeek reportedly comes from High-Flyer Capital, a hedge fund also founded by Liang in 2015 after his graduation from Zhejiang University. Before focusing on AI, he worked in software development and later used AI-based systems to analyze financial markets, significantly increasing the fund’s success.

By 2021, High-Flyer Capital’s assets had reportedly exceeded 100 billion yuan. Around that time, reports emerged that Liang had purchased a large number of Nvidia GPUs, including around 10,000 chips, which later proved to be a strategic advantage as U.S. export restrictions tightened.

In 2022 and 2023, the United States imposed restrictions on the export of advanced AI chips to China, including Nvidia’s high-performance H100 and later the modified H800 chips. These measures were intended to slow China’s rapid progress in artificial intelligence development.

Despite these limitations, DeepSeek has continued to advance. Liang has acknowledged that access to high-end computing hardware remains one of the company’s biggest challenges, even more than funding.

Analysts say DeepSeek’s progress is particularly notable given the technological restrictions and intense global competition in AI. While U.S. companies maintain an advantage in cutting-edge chip technology and research infrastructure, China’s rapid advancements highlight a narrowing gap in certain AI capabilities.