Zhipu AI has introduced GLM-5.2: 1 million token context and open-source code for agent tasks
Chinese tech startup Zhipu AI has officially launched its new flagship language model — GLM-5.2. This solution is designed for long-duration agent scenarios and programming tasks. The model is open-source, supports a context window of up to 1 million tokens, and is licensed under the MIT scheme, allowing local deployment.
According to the model card on the Hugging Face platform, GLM-5.2 is intended for text generation in English and Chinese. The number of parameters is an impressive 753 billion. This makes it one of the largest open-source models to date.
Architectural Innovations and Performance
A key feature of GLM-5.2 is support for multiple levels of "reasoning intensity." This allows users to flexibly balance between response quality and latency time. The architecture also integrates IndexShare mechanisms and an updated MTP layer for speculative decoding.
Developers claim that IndexShare reuses one indexer for every four layers of sparse attention, reducing operations per token by 2.9 times. The MTP update, in turn, increases confirmation length by 20%. These optimizations are critical for handling long contexts.
Benchmarks and Comparison with Competitors
According to test results 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 a leading position among open-source counterparts.
The model is available for local deployment via SGLang, vLLM, Transformers, KTransformers, and Docker Model Runner. Quantizations for llama.cpp, Ollama, and LM Studio are also supported, making GLM-5.2 accessible to a wide range of developers and researchers.
Expert opinion: The release of GLM-5.2 is not just another step in the AI race, but a clear signal of a shift in focus toward practical applicability. The ability for local deployment and the open MIT license make this model extremely attractive for the enterprise sector, especially amid growing data privacy requirements. 753 billion parameters with a 1 million token context is a powerful tool for automating complex multi-stage processes.