Kingsoft Cloud announced that its StarFlow training and inference platform has fully integrated the Zhipu GLM series of inference models, becoming one of the first platforms to integrate this series. This move marks Kingsoft Cloud's further expansion in the AI field, providing users with more efficient, intelligent, and cost-effective model services.
The Zhipu GLM series of inference models, recently open-sourced by Zhipu, a leading Chinese AI unicorn, includes base, inference, and reasoning models, all under the MIT license. Among them, GLM-Z1-32B-0414 is a reasoning model with deep thinking capabilities, surpassing its 32B base model in mathematical abilities and complex task solving. In benchmark tests like AIME24/25, LiveCodeBench, and GPQA, GLM-Z1-32B-0414 excels in mathematical reasoning, instruction following, and comprehensive tool usage. Additionally, Zhipu has trained a smaller 9B parameter model, which, despite its smaller size, still demonstrates excellent performance in mathematical reasoning and general tasks.
Kingsoft Cloud's StarFlow training and inference platform focuses on efficient deployment and distributed inference collaboration, providing services from model deployment to invocation. Through intelligent task scheduling and cross-node parallel computing, the platform achieves efficient model inference. Users can build stable and cost-effective model services in a public cloud environment without complex configurations. With the integration of the GLM series inference models, the deployment process on the Kingsoft Cloud StarFlow platform is further simplified. Users can create resource pools, navigate to the model online service page, create new online model services, and configure basic information, resources, and network settings such as the number of service replicas. After deployment, users can invoke model services via intranet, internet, online debugging, or through a WebUI.
Previously, the Kingsoft Cloud StarFlow training and inference platform has deployed multiple large models, including DeepSeek. By integrating more high-quality large models, Kingsoft Cloud will provide more efficient, intelligent, and cost-effective model services to customers across various industries. This move not only enhances Kingsoft Cloud's service capabilities in the AI field but also provides more choices for developers and enterprise users, promoting the widespread application and development of AI technology.