Moore's thread MTT S5000 pioneered the adaptation of GLM-5.

On February 11th, Zhigu officially released the new generation large model GLM-5. Based on the SGLang reasoning framework, Moore Thread completed the full process adaptation and verification on the flagship AI training and reasoning integrated full-function GPU MTT S5000 on Day-0. With the wide coverage of operators and strong ecological compatibility of the MUSA architecture, Moore Thread successfully connected the entire model reasoning chain and deeply unleashed the native FP8 acceleration capability of MTT S5000, significantly reducing memory usage while ensuring model accuracy, achieving high-performance reasoning for GLM-5. This rapid adaptation not only confirmed the maturity of the MUSA software stack, but also fully demonstrated the domestic full-function GPU's immediate and efficient support capabilities for the latest large models.