GLM-5.2: China's AI Challenges Claude in Coding & Context

A new Chinese large language model, GLM-5.2, is drawing attention in technical circles. It comes from Zhipu AI and focuses on long-context reasoning and coding. What's interesting is its ability to handle extremely long inputs, like entire codebases, in a single session. This "memory span" is a key engineering focus. GLM-5.2 also introduces architectural shortcuts to manage the computational cost of these long contexts. It offers two reasoning modes: GLM-5.2 (max) for peak capability and GLM-5.2 (high) for balanced performance. The model has shown gains in coding performance and agentic task execution, with a 1 million-token context window. It's also been released with MIT-licensed open weights, and its API pricing remains unchanged from an earlier version. The bottom line is that this model is being positioned as a serious step forward, challenging established AI leaders. This could shift the landscape of advanced AI capabilities.