Chinese startup Zhipu AI has released its flagship language model GLM-5.2, designed for long agent tasks and programming. This open-source solution features a context window of 1 million tokens, an MIT license, and support for local deployment.
According to its Hugging Face profile, the model is capable of generating text in both English and Chinese and has 753 billion parameters.
GLM-5.2 supports multiple levels of “reasoning intensity,” allowing users to balance quality and latency. Its architecture includes IndexShare and an updated MTP layer for speculative decoding.
Developers claim that IndexShare reuses a single indexer for every four layers of sparse attention, reducing operations per token by 2.9 times. The MTP update increases confirmation length by up to 20%.
In three key benchmarks—FrontierSWE, PostTrainBench, and SWE-Marathon—GLM-5.2 outperformed other open-source models.
In standard programming performance tests, GLM-5.2 also emerged as the most powerful open-source model.
GLM-5.2 is distributed under the open MIT license. It supports local deployment with SGLang, vLLM, Transformers, KTransformers, and Docker Model Runner. Quantization is available for llama.cpp, Ollama, and LM Studio.
Earlier, in June, the Rio de Janeiro IT company IplanRIO introduced Rio 3.5 Open 397B as an open AI model funded by the government. However, the next day, the Nex team claimed the tool appeared to be a direct merger of Nex-N2-Pro and Qwen3.5-397B-A17B.
