Chinese AI giant Zhipu AI releases GLM-5.2: 1 million token context and open-source code
Chinese startup Zhipu AI has unveiled its new flagship language model — GLM-5.2. This solution is designed for long agentic scenarios and programming tasks. The model is distributed as open source under the MIT license, making it accessible to a wide range of developers and researchers.
The main feature of GLM-5.2 is a context window of 1 million tokens. This allows the model to process vast amounts of information, which is critical for complex analytical tasks and working with large documents. On the Hugging Face platform, the model is listed as a tool for text generation in English and Chinese. Its size is impressive — 753 billion parameters.
Architectural Innovations and Performance
GLM-5.2 offers a flexible "reasoning intensity" system, allowing users to choose a balance between response quality and latency. Two key components are integrated into the architecture: IndexShare 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. Meanwhile, the MTP update increases confirmation length by up to 20%.
In the FrontierSWE, PostTrainBench, and SWE-Marathon benchmarks, GLM-5.2 outperformed all other open-source models. In standard programming performance tests, it also took leading positions among open-source solutions.
Availability and Deployment
The model is available for local deployment via SGLang, vLLM, Transformers, KTransformers, and Docker Model Runner. For users working with lightweight frameworks, quantizations for llama.cpp, Ollama, and LM Studio are provided. This makes GLM-5.2 a flexible tool for both server solutions and personal projects.
My expert analysis: The release of GLM-5.2 is a significant step forward for open-source AI. The 1 million token context and architectural optimizations make this model competitive not only among open solutions but also in comparison with proprietary giants. However, it is important to note that previous attempts to create "open" models with state funding (as in the case of Rio 3.5) sparked debates about plagiarism. Zhipu AI appears to be taking a more transparent path, but the market still needs clear standards for verifying the originality of open-source models.