Chinese AI Model Kimi K3 Pauses Subscriptions Amidst Overwhelming Demand
The powerful AI model, a significant player in the U.S.-China tech race, suspended new sign-ups within days of its launch due to demand exceeding capacity.
The Chinese artificial intelligence model Kimi K3 has temporarily halted new subscriptions following a surge in demand that quickly overwhelmed its capacity.
Moonshot AI, the Beijing-based developer of Kimi K3, announced the pause on Sunday, stating that demand had pushed the system close to its limits within 48 hours of its public rollout. The company indicated it is prioritizing existing subscribers and is working to increase its capacity, planning to reopen subscriptions in batches.
This development underscores the challenges Chinese AI models face in meeting growing user bases both domestically and internationally, amid an intensifying technology competition between the U.S. and China. Kimi K3, with 2.8 trillion parameters, is considered the world's largest open-source AI model and was launched with the aim of competing with advanced models from companies like Anthropic and OpenAI.
"New model releases generally trigger massive interest, which can strain existing compute infrastructure," said Lian Jye Su, chief analyst at Omdia. Su suggested that Moonshot AI might not have fully anticipated Kimi K3's popularity, noting that its significant compute requirements can make allocation challenging and costly. He added that the surge in demand likely strained existing compute infrastructure.
Kimi K3 has garnered global attention and has been noted for its performance. It topped the rankings on Arena, an AI model evaluation platform, for front-end coding capability shortly after its release. The model's emergence has also raised concerns among U.S. technology companies about potential price undercutting by more affordable Chinese AI alternatives, even with U.S. restrictions limiting China's access to advanced technologies, including cutting-edge chips.
Kimi K3 is part of a wave of new, powerful Chinese AI models gaining global traction. Earlier in 2025, DeepSeek's latest V4 model also drew international attention, positioning China as a notable competitor in the AI landscape. Last month, Zhipu AI released its GLM-5.2 model, which saw rapid worldwide adoption. Additionally, Alibaba previewed its Qwen3.8 Max AI model, boasting 2.4 trillion parameters and recognized as one of the world's most powerful models, second only to Anthropic's Fable 5.