China’s AI Moonshot introduces the Kimi model, which threatens American leadership

The battle between China and the United States for supremacy in artificial intelligence escalated on Friday when Chinese start-up Moonshot AI released a new artificial intelligence model that appeared to narrow the gap on well-funded American competitors.

Moonshot said the Kimi K3 model is the world’s largest open-source artificial intelligence system, allowing anyone to freely use, modify and build it. The company said that Kimi K3 performed as well as leading models from OpenAI and Anthropic in some key tasks.

The release coincided with a speech by Chinese leader Xi Jinping in which he outlined an ambitious vision for the global development of artificial intelligence, making China a champion of open access to the technology.

“The development of artificial intelligence should not be a solo performance by one country, but a symphony of international cooperation,” Mr Xi said. China was “ready to cooperate with all parties to seize the opportunities of AI development,” he said.

The market reaction was immediate. Stocks fell on Friday, with the Nasdaq down about 1 percent, as investors sold shares in U.S. companies that make the computer chips needed to power AI, including Nvidia and Intel.

The development underscored the geopolitical stakes of the US-China struggle for leadership in what many see as the defining technology race of our time. President Trump has repeatedly called maintaining America’s lead in artificial intelligence a strategic imperative to preserve the nation’s economic and national security.

The emergence of artificial intelligence as a core technology with far-reaching economic and military implications has prompted the United States to slow China’s progress by limiting its access to the advanced chips needed to train advanced artificial intelligence systems.

The release of a free, open-source model of Chinese artificial intelligence that could rival the performance of expensive and computationally intensive systems by Silicon Valley’s best-funded companies has reignited concerns about whether the industry’s enormous spending on data centers is justified.

Over the past 18 months, Chinese tech companies have released a steady stream of open-source AI models that approach the performance of their American competitors, while claiming they require less computing resources. This combination of high performance and lower cost makes Chinese AI systems the preferred choice for many developers around the world.

The latest version of Moonshot adds to that momentum, raising the prospect that the Chinese company could soon match — or even surpass — the cutting-edge models created by leading U.S. artificial intelligence firms, which have raised hundreds of billions of dollars to build models that are closed and expensive to run.

According to benchmarks run by Vals AI, an independent company that evaluates the performance of AI models, the Kimi K3 ranks just below Anthropic’s Fable 5, while also outperforming OpenAI’s flagship GPT-5.6 Sol, said Rayan Krishnan, CEO of Vals AI.

Anthropic’s latest AI models have become so powerful that their capabilities – including identifying vulnerabilities in critical infrastructure such as banking networks and power grids – have alarmed governments around the world.

Graham Webster, a Stanford University professor who studies China’s open-source artificial intelligence ecosystem, said it doesn’t matter whether the model is at the absolute top because the performance gap between the leading American and Chinese models has narrowed.

“Everyone has a different number on how many months China is behind the US, and estimates tend to cluster around six months,” he said. “That’s not much of a clue.

Moonshot said they will release full details on the Kimi K3 at the end of the month.

Samm Sacks, a senior fellow at the Johns Hopkins School of Advanced International Studies who attended an AI conference in Shanghai on Friday where Mr. Xi spoke, said many experts believe it is only a matter of time before the Chinese company releases a freely available model that matches Anthropic’s most advanced systems. Such a breakthrough, Ms. Sacks said, would raise difficult questions for regulators trying to contain powerful AI technology once it is released.

Last month, another Chinese start-up, Z.ai, released a model called the GLM-5.2 that was nearly as powerful as Anthropic’s Fable 5, the leading US system. And unlike the Fable, this Chinese model was widely available.

Many software developers and start-ups in Silicon Valley quickly adopted this model, mainly because it was much cheaper than the leading US systems. It arrived just as American businesses realized they needed to find ways to reduce their AI spending.

Kimi K3 is even more powerful than GLM-5.2, according to benchmarks performed by Vals AI. The model comes as the Trump administration considers regulating the technology. Many Silicon Valley executives have begun to support the regulation, arguing that cutting-edge technology can enable malicious cyberattacks and potentially aid in the production of biological weapons.

Moonshot AI has been widely adopted in China since the Beijing-based startup was founded in 2023.

Almost all of China’s leading AI systems are open source, which helps them gain users around the world. But unlike the closed systems favored in Silicon Valley, Chinese models have proven harder to profit from. The challenge to use global adoption to create a profitable business is now fracturing some of China’s leading AI research labs, including internet giant Alibaba.

Chinese AI researchers routinely cite a shortage of computing power — caused by U.S. export controls — as the industry’s biggest constraint, and companies have spent heavily to secure shortages of chips. Some companies, including MiniMax and Z.ai, have gone public to raise money from investors. Alibaba is increasingly steering customers toward models that require users to pay.

However, the scale of fundraising by Chinese AI firms pales in comparison to their American competitors. In May, Moonshot raised $2 billion in a funding round backed by China Mobile, a state-owned telecommunications company, and the investment arm of food delivery giant Meituan. By comparison, Anthropic raised $65 billion in the same month.

Instead, Moonshot focused on making the most of its limited computing resources. Speaking at the World Economic Forum in Davos this year, company president Yutong Zhang said efficiency was a must from the start.

“We knew we didn’t have the luxury to just increase computing power,” she said.

Anthropic and OpenAI have accused the Chinese companies of improperly extracting data from their AI systems to speed up the development of competing models. But experts say Moonshot and Z.ai’s recent progress can’t be explained by distillation alone. Years of operating under export controls forced Chinese companies to innovate within tight computer constraints.

“The Chinese models are not superior just because of the distillation,” said Stanford’s Mr. Webster. “There’s real innovation going on.”

Xinyun Wu contributed research from Taipei, Taiwan.