Zhipu’s GLM-5.2 narrows the coding gap with closed models

Zhipu AI has released GLM-5.2 under the MIT license, The Decoder reports. The model offers a stable 1-million-token context and performs strongly on FrontierSWE, a benchmark for long-running coding tasks.

The release matters because open models are pressing closer to closed-source leaders in practical coding workflows. Long-context stability is especially important for agents working across large repositories and multi-hour tasks.

The model still trails closed systems on some reasoning measures, but its coding results add pressure to the economics and openness of AI developer tools.