The release of Kimi K3, an artificial intelligence model created by a former computer science graduate student at Carnegie Mellon University, has shaken parts of social media and AI commentators. The model, developed by Moonshot AI, a Chinese lab co-founded by former student Yang Zhilin, has captivated many observers with its capabilities.
“Kimi K3 demonstrates that the United States’ moat in creating cutting-edge AI software is not as durable as many of us expected,” says Ryan Fedasiuk, a fellow at the American Enterprise Institute. “We should expect Chinese AI labs to continue freely distilling and releasing a version of the US frontier at a rate just a few weeks behind US labs.”
Social media has been ablaze with stunning demonstrations of the model’s prowess, including its ability to create a compelling browser-based version of macOS in a matter of minutes. The demonstrations have led some observers to conclude that Chinese AI models have caught up to their American counterparts.
But Yang and Moonshot himself recognize what K3 is not. It’s no Claude Fable or GPT-5.6 Killer. The company’s CEO even added a line to the model’s release notes acknowledging a “notable gap” between the K3 and leading American models.
However, that gap is closing. The UK’s AI Safety Institute published a report last week suggesting that open-weight models, such as those emerging from China, are now at most four to seven months behind the frontier, compared with a lag of six to 10 months for most of 2025.
However, the gap remains and became evident on the first day that K3 was available to the public. “K3 is an extremely large and computationally demanding model. With almost 3 trillion parameters, it requires almost an entire server rack of high-end AI chips to run,” says Fedasiuk. “Moonshot had serious uptime issues on K3 launch day: the service was down and causing errors for more than 60% of users who logged into Kimi’s web portal, probably because Moonshot didn’t have enough compute access.”
Computing infrastructure is one area where China and its leading AI companies may continue to lag behind the United States. “As models become even larger and more performant, the real race is to build the computing infrastructure needed to run and deliver AI to billions of weekly active users,” Fedasiuk says.
Of note, during the same week, Kimi struggled to meet its initial demand, OpenAI reset speed limits for Codex and ChatGPT users twice, while Anthropic offered a reset of its own, even as both companies provided access to more powerful models.
K3’s apparent success has raised questions about whether Trump-era export controls designed to restrict China’s access to GPUs, the chips needed to train huge AI models to a world-leading standard, have backfired in catalyzing AI development in China. Experts say no.
“People often confuse that export controls have some disadvantages, which export control hawks agree with, as promoting efficiency, while export controls are downright harmful,” says Miles Brundage, an AI policy researcher. “There is every reason to think that China would have stronger AI capabilities without export controls and therefore more computing,” he says.
China’s success has come despite US export controls as well as comparatively weak enforcement in some areas.
K3, however, demonstrates that it is still possible to approach the technological frontier outside of American control, which seems increasingly capricious. For countries worried about incurring the wrath of Donald Trump and losing access to leading models like Mythos and GPT-5.6, as has happened before (albeit temporarily), China may seem like a tempting alternative.
Last week, Chinese President Xi Jinping spoke at the World AI Conference in Beijing. As part of a four-point plan for AI development around the world, Xi said “we should uphold openness and win-win cooperation to boost innovative development.”
Kyle Chan, a researcher at the Brookings Institution, has compared the competition to the Apple-Android battles of previous generations. “America is Apple and China is Android. That’s Beijing’s view of the world,” he wrote on
In reality, Chan said, that framing is inaccurate. China has for years benefited from the openness of American AI research, an approach that American companies are only now reconsidering as their models become more capable, potentially more dangerous and riskier to launch without restrictions.
The rest of the world may not recognize that nuance or remember the history of American openness. Instead, you can see the United States becoming increasingly determined to rationalize access to its most advanced technology and then look to China, which offers models that are almost as capable, much cheaper and accompanied by fewer apparent strings attached.
K3 doesn’t make Beijing the new center of the AI universe overnight. But it strengthens China’s gravitational pull in a way that DeepSeek ultimately did not.
