China AI Glossary
Key terms for understanding China’s AI ecosystem — from model architectures to regulators. This glossary grows with our coverage.
AI Agent
Software that uses an AI model to plan and execute multi-step tasks autonomously — browsing, writing code, or operating other tools — rather than just answering a single prompt. Agents became the defining product category of Chinese consumer AI in 2025/26.
Algorithm Registry
China’s filing system requiring companies to register recommendation and generative AI algorithms with the Cyberspace Administration of China (CAC) before public deployment. A precondition for launching any consumer-facing AI service in China.
CAC (Cyberspace Administration of China)
China’s central internet regulator and the most important authority for AI governance. The CAC drafts and enforces the core AI rules, runs the algorithm registry, and approves public model launches.
DeepSeek
Hangzhou-based AI lab, spun out of the hedge fund High-Flyer, that shook the global industry in early 2025 with V3 and R1 — frontier-class open-weight models trained at a fraction of Western costs.
Distillation
Training a smaller model to imitate a larger one, preserving much of its capability at far lower cost. Widely used across the Chinese ecosystem to build efficient models — and a recurring point of dispute with Western labs.
Doubao
ByteDance’s consumer AI assistant and model family — one of China’s most-used chatbots, distributed through ByteDance’s enormous app ecosystem.
Ernie (文心)
Baidu’s foundation model family and chatbot, one of the earliest Chinese answers to ChatGPT. See our analysis Baidu Ernie vs. Tencent Hunyuan.
Foundation Model (大模型)
A large AI model trained on broad data that can be adapted to many tasks. The Chinese term dà móxíng („big model“) became the catch-all label for the entire industry.
GLM
The model family of Beijing-based Zhipu AI, one of China’s leading enterprise-focused AI labs with strong open-weight releases.
Hundred Model War (百模大战)
Chinese shorthand for the 2023–24 period when well over a hundred companies launched their own foundation models. The field has since consolidated around a handful of frontier labs and the big platforms.
Hunyuan (混元)
Tencent’s foundation model family, distributed across WeChat and Tencent’s platform empire rather than as a standalone product.
Interim Measures for Generative AI
China’s core generative AI regulation, in force since August 2023. It requires security assessments, content moderation, data-quality controls, and labeling of AI-generated content for public-facing services.
Kimi
The assistant and model family of Moonshot AI, a Beijing lab known for long-context models and the strong open-weight K2 series.
LLM (Large Language Model)
An AI model trained on vast amounts of text to understand and generate language — the technology behind chatbots like Qwen, Ernie, Kimi, and ChatGPT.
MIIT
China’s Ministry of Industry and Information Technology — responsible for industrial AI policy, telecoms, and the semiconductor supply chain, alongside the CAC’s content-focused role.
MiniMax (稀宇科技)
Shanghai-based AI company behind strong multimodal and video models — one of the most consumer-oriented Chinese frontier labs. See our company profile.
Mixture of Experts (MoE)
A model architecture that activates only a subset of its parameters per request, drastically cutting compute costs. Pioneered at scale by Chinese labs like DeepSeek and now the dominant architecture for large models.
Open Weights
Models whose trained parameters are published for anyone to download, run, and fine-tune. China’s frontier labs made open weights their signature strategy — a key difference from the closed US frontier labs.
Qwen (通义千问)
Alibaba’s model family — the world’s most widely adopted open-weight model line, with thousands of community fine-tunes built on top of it.
Reasoning Model
A model trained to „think“ step by step before answering, trading response time for accuracy on hard problems. DeepSeek R1 made reasoning models a global open-source phenomenon.
Six Little Dragons (六小龙)
Media nickname for the group of Chinese AI startups — including Zhipu, Moonshot, and MiniMax — that emerged as frontier-model challengers to the big platforms.
SMIC
Semiconductor Manufacturing International Corporation — China’s largest chip foundry and the centerpiece of its effort to build a domestic AI chip supply chain under US export controls.
STAR Market
Shanghai’s Nasdaq-style tech board, the main domestic listing venue for Chinese AI and semiconductor companies.
Token
The basic unit of text an AI model processes — roughly a word fragment. Model pricing, context windows, and usage statistics are all measured in tokens.