GLM-5 is Zhipu AI's latest large language model, designed for complex systems engineering and long-horizon agent tasks. Released on February 11, 2026, it is a flagship open-weights model. GLM-5 is a Mixture-of-Experts (MoE) model with 744 billion parameters, of which 40 billion are active during inference. This architecture allows the model to maintain knowledge depth while managing computational costs. It integrates DeepSeek Sparse Attention (DSA) to reduce deployment costs while preserving long-context capacity. GLM-5 was trained on 28.5 trillion tokens and supports a context window of 200,000 tokens.
GLM-5 excels in coding, logical reasoning, and autonomous agent systems, demonstrating state-of-the-art performance in open-source benchmarks. It approaches the capabilities of Claude Opus 4.5 in code-logic density and systems engineering. The model is available through the Z.ai API and platforms like OpenRouter and Ajelix. GLM-5's pricing starts at $1.00 per million input tokens and $3.20 per million output tokens. Zhipu AI, the company behind GLM-5, was founded in 2019 as a spinoff from Tsinghua University and is the first publicly traded foundation model company. GLM-5 was trained on Huawei Ascend chips, marking a step towards AI infrastructure independence.
