At the Kingsoft Cloud annual Tech Talk held today, Senior Vice President Liu Tao announced that the intelligent computing platform, Kingsoft Cloud Xingliu, has completed a strategic upgrade from a resource management platform to a comprehensive, one-stop platform for the entire AI training and inference workflow. Spanning from training and inference platforms to robotics platforms and model API services, the upgraded Kingsoft Cloud Xingliu platform establishes a full-chain闭环, covering everything from heterogeneous resource scheduling and self-healing for training task failures to support for robotics industry applications and the commercialization of model API services.
Reportedly, in terms of platform efficiency, the Kingsoft Cloud Xingliu training and inference platform provides complete lifecycle management for models, from development and training to inference. It possesses capabilities across four core modules: development, training, inference, and data processing. By reducing the complexity of multi-module coordination, it enables an "out-of-the-box" AI development experience. The self-developed GPU fault self-healing technology, combined with task observability design, allows for real-time monitoring of hardware health and task progress, automatically triggering fault migration and task rescheduling to minimize the risk of computing interruptions and ensure the stable operation of long-cycle training tasks.
As a full-chain cloud-native platform designed for robotics development and deployment, the Kingsoft Cloud Xingliu robotics platform deeply integrates key stages such as data collection, storage, annotation, model development, training, deployment, and simulation. It creates a dedicated integrated engine for data, models, and simulation within embodied scenarios. For large language model application developers and enterprise users, the Kingsoft Cloud Xingliu platform's model API service offers highly available and easily integrable model invocation and management capabilities, covering the entire lifecycle of model calls. This service supports high-concurrency inference and multi-model management, helping users efficiently access various model resources and facilitating the deployment of large model applications. Currently, the model API service of the Kingsoft Cloud Xingliu platform has already accumulated numerous industry clients.
Furthermore, the model ecosystem of the Kingsoft Cloud Xingliu platform is continuously expanding. The platform currently supports nearly 40 different models, including DeepSeek, Xiaomi MiMo, Qwen3, and Kimi. Through one-stop access, customers can efficiently utilize a variety of models, enabling them to enjoy stable and efficient cloud services while focusing more on AI business innovation and value creation.
Comments