China Loses LLM Race but Can Win in AI, Ex-Tencent Lead Says

China is losing the large language model (LLM) race but can still succeed in artificial intelligence, according to Liu, former head of Tencent's AI efforts. Liu left Tencent in late 2024 after over eight years, shortly after the launch of Hunyuan, sparking speculation. In an interview with the South China Morning Post, Liu emphasized the lack of "paradigm" innovations—technical breakthroughs defining new AI eras, such as OpenAI's ChatGPT or Anthropic's Claude Code—as China's biggest weakness. He stated that Chinese companies are either copying DeepSeek or US firms at the core technical level. Despite narrowing benchmark scores, Liu noted that public metrics do not reflect real-world usefulness gaps. While US leaders like Anthropic launched the Mythos model in April, Chinese leader DeepSeek's latest V4 model failed to match previous heights.