When Chinese startup DeepSeek released details of one of its artificial intelligence models last year, it sent shockwaves through the tech industry.
The company said it built its system spending far less on computer chips than U.S. competitors like OpenAI and Anthropic. This marked the start of what became known as the “DeepSeek moment” in China, shorthand for the belief that Chinese AI companies were ready to show off their technical capabilities to the world.
The DeepSeek moment reflects a shift in the global AI landscape. The change was not just about cost reduction, but also about opening up the way technology is shared.
DeepSeek has released its models as open source, meaning others can use and modify them freely. In contrast, OpenAI and Anthropic have retained their main proprietary models. The episode demonstrated that an open source system could perform almost as well as closed versions. In the months that followed, Chinese companies released dozens of other open source models. By the end of 2025, these models accounted for a significant portion of global AI use.
On Friday, DeepSeek released a preview of V4, its long-awaited follow-up model, which it intends to open source. The new model excels at writing computer code, an increasingly important skill for leading AI systems. It significantly outperformed all other open source systems in code generation, according to tests by Vals AI, a company that tracks the performance of AI technologies.
DeepSeek released its new model just days after Moonshot AI, another Chinese startup, introduced its latest open source model, Kimi 2.6. Although these systems lag behind the coding capabilities of the leading US models from Anthropic and OpenAI, the gap is narrowing.
The implications are significant. Using AI to write code is faster and allows human programmers to focus on more important problems. This also means that users can use the latest version of DeepSeek to power AI agents, which are personal digital assistants that can use other software applications on behalf of office workers, including spreadsheets, online calendars and email services.
As AI systems get better at writing computer code, they also get better at detecting security vulnerabilities in software – a skill that fundamentally changes cybersecurity. This means that tools like DeepSeek’s can be used to both attack and defend computer networks.
In all tasks, DeepSeek V4 is on par with the latest model from Moonshot. “They are basically neck and neck,” said Rayan Krishnan, chief executive of Vals AI.
In the months leading up to DeepSeek’s latest release, overseas competitors decided to preempt a new round of glowing headlines. Silicon Valley AI giants Anthropic and OpenAI said DeepSeek unfairly relied on their technology through distillation, a process in which engineers imitate a rival model by interrogating it millions of times and copying its behavior.
The competition to build the most capable AI systems has transformed into a geopolitical power struggle. While Silicon Valley executives at Anthropic and OpenAI warn that their technology would be dangerous in the hands of autocratic countries, China has invested billions to become an AI superpower, viewing the technology as a key driver of economic growth.
DeepSeek’s open source models are at the heart of this strategy. While many Western companies hold on to their most valuable models, China has embraced open source and almost all of its most successful systems are widely available.
Despite this, Chinese AI companies face major obstacles. Three U.S. administrations have imposed export controls limiting access to advanced chips needed for cutting-edge AI systems. And Silicon Valley companies continue to outspend their Chinese rivals in the race for top AI talent.
China’s push toward open source AI has become a major economic advantage at home, according to a new study by a U.S. Congressional advisory body. With few barriers to their use, the systems have spread across industries such as robotics, logistics and manufacturing. The study found that these industrial applications generate real-world data used to improve AI systems.
This approach has allowed Chinese technology companies to gain global influence, as programmers and engineers around the world adopt their systems to create new products.
From Lagos to Kuala Lumpur, developers on tight budgets are turning to Chinese open source models because they are cheaper to operate and therefore easier to experiment with. Last May, Malaysia’s Deputy Minister of Communications said the country’s sovereign AI infrastructure would be built on DeepSeek technology.
Chinese open source models accounted for about a third of global AI use last year, according to a study by OpenRouter, an AI model marketplace. DeepSeek was the most used, followed by models from Alibaba, the Chinese internet company.
This reflects a broader strategy. As Chinese companies expand overseas, making their systems open source helps them gain traction with coders by offering cheaper, more accessible tools.
“Open source is the soft power of future technology,” said Kevin Xu, US founder of Interconnected Capital, a hedge fund that invests in artificial intelligence technologies. Mr. Xu and his fund do not invest in DeepSeek.
Wei Sun, senior AI analyst at Counterpoint Research in Beijing, said DeepSeek’s success has paved the way for Chinese tech giants to make their AI systems public rather than closely monitor them.
Alibaba has since become a leader. Its Qwen model family has surpassed 1 billion downloads. ByteDance, parent company of TikTok, also shared some details about its technology after spending $11 billion on AI infrastructure in 2024.
“The AI generation of open source builders from China has arguably been the biggest AI story in 2025,” Xu said. “The progression of models, the pace of releases, and the number of AI labs that are competing but also seem to be cheering each other on have happened fast and furious, with no signs of slowing down.”
