GLM-5.2 from Zhipu AI: A new era of open-source models with a context of 1 million tokens and an MIT license
Chinese startup Zhipu AI has unveiled its flagship language model GLM-5.2, designed for long agentic tasks and programming. The open-source solution features a context window of 1 million tokens, is distributed under the MIT license, and supports local deployment.
On the Hugging Face platform, the model is described as a tool for generating text in English and Chinese. GLM-5.2 boasts an impressive 753 billion parameters, placing it alongside the largest open-source counterparts.
Architectural Innovations and Performance
GLM-5.2 offers several levels of "reasoning intensity," allowing users to flexibly balance between response quality and latency. The architecture incorporates IndexShare mechanisms and an updated MTP layer for speculative decoding. According to the developers, IndexShare reuses one 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 key benchmarks—FrontierSWE, PostTrainBench, and SWE-Marathon—GLM-5.2 outperformed all other open-source models. In standard programming performance tests, it also took the leading position among open-source solutions.
Availability and Deployment
GLM-5.2 is distributed under the open MIT license, making it accessible to a wide range of developers. For local deployment, support is claimed for popular frameworks: SGLang, vLLM, Transformers, KTransformers, and Docker Model Runner. Quantizations for llama.cpp, Ollama, and LM Studio are also available, simplifying integration into existing projects.
My expert opinion: GLM-5.2 demonstrates that open-source models are not only catching up to proprietary solutions but also surpassing them in specialized tasks such as programming and working with long contexts. The MIT license and support for local deployment make this model particularly attractive for the corporate sector, where data control is critical. In the coming months, we will likely see a wave of GLM-5.2 implementations in real business processes.