Z.ai releases GLM 5.2 to challenge US models

Z.ai has released GLM 5.2, a Chinese open source AI model it says is approaching the performance of leading US systems including Anthropic’s Claude Opus 4.8 and OpenAI’s GPT 5.5. The launch came just a day after the United States banned Anthropic from supplying its Fable 5 and Mythos models to non-Americans earlier this month, before the controls were lifted on June 30.

The company says GLM 5.2 runs with a 1 million token context window, enough to hold roughly 750,000 words in working memory, and is designed for long coding tasks. In Z.ai’s tests on long, complex coding work, it trailed Opus 4.8 by just 1%, beat GPT-5.5 and Opus 4.7 on a single-GPU model-improvement benchmark, and lagged Opus 4.8 by 13% on marathon engineering tasks while ranking second overall.

Its open source release and lack of regional limits contrast with closed-source systems from Anthropic and OpenAI, giving users the ability to modify and redistribute the model. The launch fits a broader US-China competition in which Washington has restricted semiconductor access while Chinese firms pursue cheaper open-source systems, including DeepSeek’s R1 in January last year.