Silicon Valley and Washington immediately started talking about the models, especially the K3, which is widely considered the best of the bunch. David Sacks, a venture capitalist and artificial intelligence adviser to President Donald Trump, called the operation of the Moonshot model “regarding” Earlier this week, Commerce Secretary Scott Bessent suggested the US can impose sanctions on Chinese artificial intelligence companies.
On Wednesday, Michael Kratsios, director of the White House Office of Science and Technology Policy, alleged that the Trump administration has “information that Moonshot AI has distilled Anthropic’s Fable to develop the K3 model,” which he says amounts to “stealing proprietary US technology and undermining US research” and is “unacceptable.” (Moonshot AI did not immediately respond to a request for comment.)
The new Chinese models have several things in common: third-party benchmarks show they perform almost as well as the best Western models; are optimized for agent coding tasks (the hottest thing in AI this year); and are or will soon be released with open weights, making them accessible and transparent.
But perhaps the biggest parallel between now and January 2025 – when the world was shocked by DeepSeek’s R1 model – is that it confirms that US and Chinese AI labs are taking divergent paths when it comes to being open or closed.
When it first arrived on the scene, DeepSeek challenged the assumption that only closed-source models built with billions of dollars of investment in computing infrastructure and training could achieve pioneering performance. Since then, however, Western AI labs have continued to develop AI in the same way, and today, American frontier models feel more disconnected from reality than they did a year ago.
Anthropic has claimed for months that its latest Mythos model is so dangerously good at hacking that only approved contributors can use it. When it was finally made more widely available, the White House responded by imposing broad export controls, which forced Anthropic to temporarily take Mythos and its less powerful sister model, Fable 5, offline. OpenAI similarly delayed the release of GPT 5.6 after receiving a request from the White House.
Meanwhile, in China the situation is completely different. Chinese startups and tech giants have doubled down on open-source software: anyone with a good enough computing environment can now download an open-source model, run it locally, add customizations, and generally enjoy a much greater degree of freedom than OpenAI and Anthropic would ever allow. In many ways, the open versus closed debate is more closely linked to the US-China debate than ever before.
There are many reasons why Chinese laboratories have chosen a business strategy based on open source models. Being a newer and smaller fish in the field of artificial intelligence, making their models free and open could help Chinese companies attract more users, collaborators and media attention. It also puts them on a different competitive path than that of OpenAI, Anthropic, Google, SpaceX and other wealthy giants.
Earlier this year, rumors spread that it could be Alibaba considering joining closed source race after the reorganization of teams dealing with the development of corporate artificial intelligence models. But the tech giant announced Monday that it will re-release the latest version of Qwen – a line of open-source models beloved by the global tech community – with open scales, signaling to customers and the public that it’s not turning around just yet.