China’s LLM Market Overview

Scaling Trends, AI Agents, and Emerging Technologies

As China’s AI industry accelerates, the development and deployment of large language models (LLMs) have become a central focus. From shifts in scaling paradigms to the rise of AI Agents and AI-powered hardware like smart glasses, the LLM landscape is rapidly evolving. Here’s an in-depth look at the trends and breakthroughs shaping the sector.

Scaling Challenges and Innovations in LLMs

Global and domestic research indicates that if current trends persist, the data reserves fueling LLMs could be exhausted by 2028. This has sparked concerns about the potential plateau of model performance. However, OpenAI’s Senior VP Mark Chen refuted the “Scaling Law wall” theory, emphasizing that both OpenAI’s O-series and GPT-series models remain scalable.

In China, reports suggest that some AI “Big Six” companies have scaled back pre-training efforts, with only a few, like Zhipu AI and MiniMAX, continuing to invest.

Despite these shifts, innovations in inference-driven scaling have emerged as a promising alternative. Databricks co-founder Ion Stoica highlighted that while progress in common-sense reasoning has slowed, advancements in complex problem-solving and code generation remain robust.

Despite short-term bottlenecks, experts believe that hardware and software advancements will continue to fuel LLM evolution. The focus is shifting toward more efficient resource utilization and diversification in application development, ensuring sustainable growth.

AI Agents: The Next Frontier

2024 marked a surge in AI Agent launches, with companies like Zhipu AI introducing AutoGLM, capable of executing over 54-step tasks autonomously across apps. Similarly, OpenAI is preparing to launch its AI Agent “Operator” in 2025, signaling a commercial explosion for such technologies.

Gartner predicts that by 2028, AI Agents will autonomously handle at least 15% of daily work decisions. The Chinese market for AI Agents is projected to grow from ¥55.4 billion in 2023 to ¥852 billion by 2028, with a compound annual growth rate of 72.7%.

AI Agents are already disrupting sectors like smart home solutions, healthcare, and financial services. With enhanced capabilities for tool usage (L3) and early self-learning (L4), these agents promise to redefine human-machine interaction and improve operational efficiency.

The Rise of AI-Powered Smart Glasses

Tech giants like Baidu, Xiaomi, and Huawei have unveiled AI glasses, aiming to integrate LLM capabilities into wearable devices.

Baidu’s XiaoDu AI Glasses, for instance, are the first native Chinese AI hardware product in this category, slated for release in early 2025. Xiaomi’s upcoming AI glasses, developed in collaboration with Goertek, are projected to sell over 300,000 units.

According to IDC, global shipments of smart glasses grew by over 200% in Q1 2024. By 2029, annual sales are expected to reach 55 million units, generating ¥106.8 billion in revenue. These developments highlight the strategic importance of hardware in operationalizing AI advancements and capturing market share.