Zhipu Unveils GLM-5 Coding Agent Inference Engineering Techniques for the First Time
6 hour ago / Read about 0 minute
Author:小编   

As reported by the Sci-Tech Innovation Board Daily on April 30, Zhipu published a technical blog post in the early hours, revealing, for the first time, the advancements in the foundational inference technologies for the GLM-5 series models when applied in ultra-large-scale Coding Agent deployment scenarios. Notably, the system's throughput has witnessed a significant boost of up to 132%, accompanied by a reduction in abnormal output rates. Moreover, the repair solution has been embraced by the SGLang open-source community.