The China Internet Network Information Center (CNNIC) Policy and International Cooperation Institute recently released the Generative AI Application Development Report (2026), finding that China's pace of indigenous AI innovation is quickening, with rapid breakthroughs across multiple links of the whole industry chain.
A headline finding: on global mainstream large-model call leaderboards, the top six models all come from Chinese teams. The report says this marks China's generative-AI industry as having moved from keeping up in exploration to running abreast of - and in places leading - the global field.
From base-model research and domestic-compute adaptation to industry-scenario deployment and cloud-network infrastructure, an autonomous, controllable and coordinated AI industrial ecosystem is taking shape. Local large models such as China Telecom's Xingchen (TeleChat) are moving beyond lab tests into industrial and government scenarios on domestic software and hardware.
Telecom operators are building shared foundations: China Telecom, relying on its self-developed Xingchen model and Xirang compute-scheduling base, has built the Xingchen TokenHub platform, aggregating multiple mainstream large models and advancing integrated tuning of domestic chips and models to give developers one-stop model-calling services.
Image: Efficiency of AI-related computer chips. Source: Wikimedia Commons (CC BY-SA 4.0).
