Chinese artificial intelligence companies have formed two new industry alliances aimed at developing a domestic AI ecosystem to reduce reliance on foreign technology amid ongoing U.S. export restrictions on advanced Nvidia chipsets. In parallel, Chinese AI startup Zhipu AI released its new-generation flagship open-source large language models, GLM-4.5 and GLM-4.5 Air. The GLM-4.5 model features 355 billion total parameters with 32 billion active parameters, while the GLM-4.5 Air variant has 106 billion total and 12 billion active parameters. These models utilize a hybrid mixture-of-experts (MoE) architecture designed to unify advanced reasoning, coding, and agentic capabilities, supporting complex tasks with a "thinking mode" and instant response with a "non-thinking mode." The GLM-4.5 series has demonstrated competitive performance on multiple benchmarks, ranking third on a mixed leaderboard of 12 challenging tests, and is reported to match or surpass global top models like Claude 4 Opus and Gemini 2.5 Pro, while costing approximately one-tenth of Claude’s API pricing. The models have attracted significant attention on global AI platforms such as Hugging Face and X, highlighting China's growing presence in AI research despite U.S. chip export restrictions. Concurrently, Nvidia has advanced its own AI offerings with the release of Llama Nemotron Super 49B v1.5, which recently topped the Artificial Analysis Intelligence Index leaderboard in the 70 billion parameter open model category. This model is noted for best-in-class performance in reasoning and agentic tasks and is commercially available via Hugging Face. Overall, China is intensifying efforts to build a self-sufficient AI ecosystem capable of competing globally without dependence on Western technology, as underscored by recent academic research and media reports.