Amid tightening US export restrictions on AI technologies, Chinese developers are rapidly scaling up their AI models to trillion-parameter architectures to narrow the gap with US leaders like OpenAI and Anthropic. This surge signals a major shift in China’s AI strategy focusing on domestic innovation and competitive pricing.

  • Chinese AI models now exceed a trillion parameters, rivaling top US systems.
  • Export controls by the US have accelerated China's push for self-reliant AI technology.
  • Innovations in model architecture reduce costs and improve commercial viability.

What happened

Chinese AI developers have accelerated their efforts to create foundation models exceeding one trillion parameters, marking a significant leap from the billion-parameter models that prevailed in 2023 and early 2024. Major Chinese companies like DeepSeek, Xiaomi, and Alibaba have released or announced trillion-parameter models, integrating advanced computing architectures such as Mixture of Experts to enhance performance and efficiency.

This development comes as US authorities implement unprecedented export controls restricting foreign access to leading US AI software, including models from Anthropic. Consequently, Chinese firms are increasingly focused on domestically developed chip stacks and AI models, aiming to reduce dependence on US technology and catch up with global leaders like OpenAI and Anthropic.

Why it matters

The growth of China’s trillion-parameter AI models represents a strategic shift in the global AI race, underscoring mounting technological competition between China and the US. Model scale has become a critical yardstick for investors and enterprises evaluating AI capabilities, with larger models often perceived as more powerful and capable, especially for complex tasks requiring extensive contextual understanding.

Additionally, the adoption of more efficient MoE architectures has substantially lowered training and inference costs, making giant AI models more commercially sustainable. Chinese companies are also competing aggressively on pricing, offering AI services at a fraction of the cost of US counterparts, which could accelerate adoption domestically and in other markets, further strengthening China’s AI ecosystem.

What to watch next

The continued expansion of AI model parameters in China will be a key indicator of technological progress, particularly as companies push beyond trillion-parameter thresholds and refine architectures to balance scale, cost, and real-world application efficacy. Attention will also focus on deployment strategies emphasizing model quality and data refinement alongside sheer size.

Market reactions will be important to monitor, as investor confidence rises or falls based on companies’ ability to keep pace in the trillion-parameter race. The industry may also witness innovative uses of these massive models for synthetic data generation and training smaller specialized models, potentially influencing government procurement policies and enterprise adoption patterns in China’s AI sector.

Source assisted: This briefing began from a discovered source item from SCMP China Tech. Open the original source.
How SignalDesk reports: feeds and outside sources are used for discovery. Public briefings are edited to add context, buyer relevance and attribution before they are published. Read the standards

Related briefings