US dominance of the AI landscape is receding with the release of Kimi K3, the newest model developed by a Chinese start-up called Moonshot AI.
Kimi K3 is as “powerful as the most advanced AI systems” from the US, such as OpenAI and Anthropic, while being “cheaper to run”, said The Telegraph. It is available for people to freely download and is an “open-weight model”. This allows developers to download it, modify it and run it on their own servers for free, “positioning it for easier adoption and a wider market globally”, said CNN.
How good is K3?
K3 was “beating” Anthropic’s Opus 4.8 model and OpenAI’s 5.5 in most coding and “general AI agent benchmarks”, according to results released by Beijing-based start-up Moonshot. This was while costing “about a third of Anthropic’s Opus 4.8”, said the Financial Times. But K3 was “falling short” of Anthropic’s Fable model.
The Week
Escape your echo chamber. Get the facts behind the news, plus analysis from multiple perspectives.
SUBSCRIBE & SAVE
Sign up for The Week's Free Newsletters
Join more than 350,000 subscribers and keep yourself informed with a selection of The Week’s most interesting, enlightening and entertaining stories - plus daily puzzles.
Join more than 350,000 subscribers and keep yourself informed with a selection of The Week’s most interesting, enlightening and entertaining stories - plus daily puzzles.
Latest Videos FromThe Week
Unlike the “closed” and “proprietary” American systems, K3’s “open nature” allows global users to “modify the system for advanced reasoning and complex software development”, said the BBC. It will be the world’s first open source model in the “three-trillion-parameter class”. A parameter refers to “learned values within a machine learning model that determine how it maps input data into outputs”, said IBM.
The release of K3 had an “immediate” market reaction, said The New York Times. The Nasdaq dropped around 1%, with investors selling shares in US companies like Intel and Nvidia, who make computer chips to power AI.
How have Anthropic and OpenAI reacted?
Anthropic and OpenAI have accused Chinese companies of “improperly harvesting data” from their systems to “accelerate” their own competing models’ development. However, after years of “operating under export controls” and “tight computing constraints”, China’s advances “cannot be explained by distillation alone”. There is “real innovation going on”, said Graham Webster, a Stanford University professor.
The “steady stream” of Chinese open-source AI models, including K3, that “rivals” the performance of “costly systems from Silicon Valley” have become the “preferred choice” for developers, said the NYT. President Trump has “repeatedly framed maintaining America’s “lead” in the “defining technological race of our time” as a “strategic imperative for preserving the country’s economic and national security”. These technological developments have “underscored the geopolitical stakes of the US-China contest for leadership”.
What does this mean for US dominance in AI?
Xi Jinping released a statement saying AI development should not be a “solo performance by a single country” but instead a “symphony of international cooperation”. He said China was “ready to work with all parties to seize the opportunities of AI development”.
US AI leaders and policy makers had taken “comfort” from “estimates that China remained six to 12 months behind”, even with Chinese open-weight models “gaining momentum”, said Axios. However, the release of Kimi K3 suggests that “cushion may have collapsed far faster than expected”. With the “entire game changed”, this could “trigger some code red for some”, said AI analyst Kim Isenberg.
American labs are “hardly out of ammunition” with OpenAI and Anthropic “racing ahead” to build newer systems like GPT-6 and Claude Opus 5. But even if the US “pulls ahead” again, China has shown it can “close the gap quickly”.
