Summary
Asianometry profiles Zhipu AI, also known as Z.ai, the Chinese lab behind the well-regarded GLM series of open-weight language models. The company was one of two Chinese AI labs, alongside MiniMax, to list on the Hong Kong stock exchange in early 2026, and the video asks whether Zhipu is to China what Anthropic is to the West.
The story begins at Tsinghua University’s Knowledge Engineering Group, where founder Zhang Peng and professor Tang Jie worked on the AMiner research-mining platform. The video traces how the team commercialized that work under a “Paper to Product” culture, formally founding Zhipu in June 2019, and how the arrival of GPT-3 pushed the company toward large language models.
Viewers also get details on the company’s early projects, including the WuDao 2.0 mixture-of-experts model, the virtual student Hua Zhibing, and its “Model-as-a-Service” business model. The video covers the training of GLM-130B, a dense model trained on 96 A100 servers over two months at an all-in cost of roughly $1.5 million, and explains why the team chose to open-source it.
📺 Source: Asianometry · Published November 01, 2026
🏷️ Format: Deep Dive
