China: Chinese open-source LLMs undergo rapid iteration, driving global AI development.

【Voice_over】 China’s top tech companies including Tencent, DeepSeek, Moonshot AI, and MiniMax, are rapidly rolling out upgrades to their individual open-source large language models, or LLMs. Tencent's new Hunyuan model significantly reduces AI inference and deployment costs, which allows small and medium-sized enterprises to build customized AI applications without breaking the bank. DeepSeek, meanwhile, just released its highly anticipated V4 model. Creators say the model’s Pro edition is advanced in agentic coding and general knowledge, second only to Google’s Gemeni 3.1 Pro, which is closed source. Moonshot AI from Beijing introduced an upgraded Kimi model that incorporates advanced task decomposition, allowing multiple AI agents to work together on complex workflows. And from Shanghai- MiniMax has achieved notable breakthroughs in code generation and program comprehension. 【Sound_bite】 Zhong Xinlong, associate researcher, China Center for Information Industry Development (CCID): "Open-source models like DeepSeek's V4 lower enterprise adoption costs, expand the developer ecosystem, and enhance supply chain autonomy. Open-source large language models will accelerate the transition of AI from standalone tools to foundational industrial infrastructure." 【Voice_over】 According to the Spring 2026 report from Hugging Face, the world’s largest open-source AI community, Chinese-developed models accounted for about 41 percent of LLM downloads on its platform over the past year. The report names China as one of the most active and fastest-growing regions for open-source AI models in the world. [Restrictions: No access Chinese mainland]