GLM-5.2 from Zhipu AI: a context window of 1 million tokens and open-source code for agent tasks
Chinese developer Zhipu AI has unveiled its flagship language model GLM-5.2, designed for long-term agent scenarios and programming tasks. The solution is distributed as open source, features a context window of 1 million tokens, an MIT license, and supports local deployment.
According to specifications on Hugging Face, the model is intended for text generation in English and Chinese and has 753 billion parameters. GLM-5.2 offers several levels of "reasoning intensity," allowing a balance between response quality and latency.
The architecture includes an innovative IndexShare mechanism 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. In turn, the improved MTP increases confirmation length by up to 20%.
In three key benchmarks — FrontierSWE, PostTrainBench, and SWE-Marathon — GLM-5.2 surpassed all other open-source models. In standard programming performance tests, it also took leading positions among open-source solutions.
GLM-5.2 is distributed under the open MIT license. For local deployment, support is claimed for SGLang, vLLM, Transformers, KTransformers, and Docker Model Runner. Quantizations are available for llama.cpp, Ollama, and LM Studio, making the model accessible to a wide range of developers and researchers.
Analyst's opinion. The release of GLM-5.2 is a significant step in democratizing powerful language models. The 1 million token context and open MIT license create serious competition for proprietary solutions, especially in the segment of agent tasks and programming automation. However, the key challenge remains the practical implementation of local deployment with 753 billion parameters — without serious hardware resources, this will be difficult.