Chinese AI models appear close to the frontier primarily because they are trained on the outputs of leading U.S. models. This creates a dependency loop: they can only catch up by using the latest from the West, ensuring they remain followers rather than innovators who can achieve a true breakthrough.

Despite impressive models from companies like DeepSeek, China's AI ecosystem is heavily reliant on "distilling"—essentially copying and refining—open-source models from the US. This dependency on an external innovation engine is a major weakness in their national strategy to achieve genuine AI leadership and self-sufficiency.

China is gaining an efficiency edge in AI by using "distillation"—training smaller, cheaper models from larger ones. This "train the trainer" approach is much faster and challenges the capital-intensive US strategy, highlighting how inefficient and "bloated" current Western foundational models are.

Facing compute and capital shortages, Chinese AI labs don't pioneer frontier research. They wait for Western labs to publish breakthroughs, likening it to 'knowing the answer to the homework,' then work backwards to replicate them, focusing resources on efficient post-training.

China is rapidly closing the AI gap not through pure innovation but through "distillation"—systematically querying Western frontier models via their APIs to harvest their reasoning processes. This allows them to train their own models to a near-frontier level at a fraction of the cost, bypassing years of foundational research.

Chinese labs use 'smart distillation,' a sophisticated technique where a frontier model acts as a 'teacher' to guide a smaller model's judgment and data labeling. This is viewed as a legitimate and efficient catch-up method, distinct from simply copy-pasting answers.

US officials and AI labs allege Chinese firms are engaged in industrial-scale IP theft. They reportedly use fraudulent accounts to extract capabilities from US models like Claude to train their own, creating a facade of domestic innovation.

In the vacuum left by banned US frontier models, Chinese labs are releasing powerful and cost-effective open-source alternatives like ZAI's GLM 5.2. These models are proving competitive on valuable, complex tasks like UI design and coding, but at a fraction of the cost.

Leading Chinese AI models like Kimi appear to be primarily trained on the outputs of US models (a process called distillation) rather than being built from scratch. This suggests China's progress is constrained by its ability to scrape and fine-tune American APIs, indicating the U.S. still holds a significant architectural and innovation advantage in foundational AI.