Why Former Tencent AI Lead Liu Wei Says China Is Losing the LLM Race — and How It Could Still Win AI

Artificial intelligence researcher Liu Wei, formerly a distinguished scientist at Tencent and a technical leader of its Hunyuan large‑model team, has offered a blunt assessment of the global AI race: China is currently behind in large language models (LLMs). But in his view, that does not mean the broader AI competition is already decided.

His argument centers on a key idea: the difference between following a technological paradigm and creating one. According to Liu, countries or companies that merely copy the current frontier risk falling further behind whenever the leaders redefine it.

Why Liu Wei Says China Is Losing the LLM Race

Liu’s core explanation is strategic rather than purely technical. He argues that much of China’s AI development has focused on replicating or catching up to models already pioneered elsewhere, such as the ChatGPT‑style conversational assistant.

The problem with this approach, he says, is structural. If a country is primarily following the current frontier, the leaders can shift the field again with a new breakthrough, leaving followers scrambling to catch up.

Liu described this risk bluntly: in the AI era, being a technological follower means the "rug can be pulled out from under you" whenever the frontier moves.

In other words, China’s challenge is not just that its models might trail the latest releases. The deeper issue is that the direction of the industry is still being set elsewhere.

What Liu Means by a New AI “Paradigm”

When Liu talks about a new AI “paradigm” (the term he repeatedly used in interviews), he is referring to a breakthrough that reshapes how AI is built, used, and commercialized, rather than incremental improvements to existing models.

Examples of paradigm‑level shifts in recent years include:

• ChatGPT, which transformed language models into mainstream conversational tools used by hundreds of millions of people.
• AI coding systems such as Claude Code, which redefined how developers interact with software tools.

These breakthroughs did more than improve accuracy—they changed workflows and user behavior, creating new industry centers of gravity.

From Liu’s perspective, competing by making slightly stronger chatbots is not enough. Instead, success requires creating the next category‑defining advance that reorganizes the field itself.

Why He Left Tencent’s Hunyuan Team

Liu spent years as a key figure in Tencent’s AI efforts, helping lead development of the company’s Hunyuan large language model. But in late 2024 he left Tencent after more than eight years with the company.

He subsequently founded Video Rebirth, an AI startup focused on video generation and so‑called “world models.”

The company is building AI systems designed to generate high‑fidelity video with strong physical consistency and control for professional uses such as advertising, e‑commerce, film, and animation.

Video Rebirth raised tens of millions of dollars in early funding and plans to develop models aimed at professional creators rather than consumer novelty tools.

The direction of the company suggests Liu believes multimodal and video‑based world models could represent a new frontier in AI, potentially forming the foundation for the next paradigm shift.

How China Could Still Win the Broader AI Competition

Despite his criticism of China’s current position in LLMs, Liu does not believe the overall AI race is settled.

His argument is that AI progress happens in waves of paradigms, and leadership can shift when the next major breakthrough arrives.

That means the key strategic question is not simply:

• Who has the best chatbot today?

Instead, it is:

• Who invents the next way people use AI?

If China can originate a new paradigm—whether in agents, multimodal systems, world models, or another emerging direction—it could redefine the competitive landscape rather than playing catch‑up within it.

In Liu’s framing, losing the current LLM phase does not necessarily mean losing the future of AI. The decisive factor will be who creates the next breakthrough that reshapes the industry.

The Bigger Picture

Liu Wei’s perspective reflects a broader debate within the global AI community. Some analysts argue the United States still holds a lead in frontier models and infrastructure, while China has moved quickly to scale deployment and develop its own model ecosystem.

His warning highlights a strategic tension common in technology races: catching up to the present can be less important than inventing the future.

For Liu, the real opportunity lies not in building a slightly better version of today’s large language models—but in discovering the next paradigm that makes them look outdated.