China's Kimi K3 pauses new subscriptions as overwhelming demand strains capacity

Kimi K3, the Chinese artificial intelligence model, has caused quite a stir in the US tech industry. Amid surging demand for the model, Moonshot AI has temporarily suspended new subscription requests after a flood of users overwhelmed its computing capacity just days after launch.

The development highlights the challenges Chinese AI companies face in serving a rapidly growing user base at home and abroad as the AI race between the US and China intensifies. Chinese open-source, lower-cost advanced AI models, such as DeepSeek's latest V4, have been gaining global attention. Earlier in 2025, DeepSeek shook global markets and helped establish China as a serious competitor to the US in frontier AI. Unlike many leading US models, Chinese frontier AI models are largely open source, making them accessible for developers and researchers to examine, modify and build upon.

Following the sudden surge in subscribers, Moonshot AI said it is prioritising existing users and will temporarily pause adding new subscribers. The company said, "We're adding capacity as fast as we can and will reopen new subscription spots in batches."

K3's arrival has rattled companies such as Anthropic and OpenAI. With 2.8 trillion parameters—a key measure of an AI model's capabilities—it is regarded as the world's largest open-source AI model.

More from Tech

Moonshot AI said it underestimated the strong demand for Kimi K3, whose high computing requirements made infrastructure allocation both difficult and costly. Since its launch, the model has climbed to the top of Arena's front-end coding rankings and has heightened concerns that lower-cost Chinese AI models could challenge the pricing power and demand enjoyed by major US AI companies.

Kimi K3 and its capabilites 

Moonshot has launched Kimi K3 with a host of bold claims about its capabilities. According to the launch notes, Kimi K3 contains 2.8 trillion parameters, a key measure of an AI model's capabilities. One of its biggest strengths is its open-source nature, which allows developers and users worldwide to examine, modify and build on the model for advanced reasoning and complex software development tasks.

According to independent benchmarks, Kimi K3 ranked first in web interface engineering and outperformed Anthropic's Fable model in blind human preference tests.