SenseCore Awarded Industry’s First 5A Native AI Cloud Platform Certification

Recently, SenseTime’s large-scale system SenseCore, a native AI cloud platform, passed the authoritative “Computing, Model, Data, Application — Computing Platform Service Capabilities” evaluation conducted by the China Academy of Information and Communications Technology (CAICT) and the Telecommunication Technology Labs (TTL). It received the industry’s highest 5A Excellence Certification, becoming the industry’s first native AI cloud platform to obtain 5A certification.

The evaluation covered core dimensions including computing power operations, management, scheduling, monitoring, and enablement. SenseCore demonstrated outstanding performance across all indicators, particularly in large-scale computing resource management, high-performance scheduling, elastic fault tolerance, heterogeneous multi-chip adaptation, and integrated training–inference capabilities. This fully showcased the system-level strength and maturity of a “Native AI Cloud Platform.”

Comprehensive Leadership: Building the “Native AI Cloud Foundation” for the Large Model Era 

End-to-End Commercial Operations Capability  

SenseCore has established a complete commercial operations system covering user management, product management, account systems, and operational analytics. Testing showed that product listings and delistings are instantly synchronized across front and back ends, bills are automatically generated and dynamically updated with resource consumption, and billing accuracy meets financial-grade standards. Tenants can independently recharge, consume, and trace transactions, while operators can monitor supply-demand and revenue in real time, providing reliable support for large-scale computing resource circulation.

Intelligent Scheduling of Heterogeneous Computing Power, Stable Support for Mixed Training   Through intelligent strategies, the platform dynamically schedules and optimizes heterogeneous computing resources, supporting containerized tasks and unified management of multi-brand GPUs. In testing, it successfully completed an 8-replica × 8-card heterogeneous training task, achieving precise matching under multiple strategies such as GPU model, physical topology, task priority, and resource affinity. Data showed an effective training duration ratio of 99.46%. The platform also supports fault tolerance for network failures and communication libraries, with real-time saving of training states during failures, greatly enhancing large model training stability.

Unified Computing Resource Management, Broad Adaptation to Chinese Chips  

SenseCore enables unified access and management of heterogeneous resources, supporting automatic node discovery, flexible definition of computing specifications, and compatibility with multiple GPU brands. In testing, the platform successfully managed NVIDIA GPUs and multiple Chinese-made chips. Newly added nodes were automatically identified and integrated into clusters, with accurate resource specification matching during instance creation. Users can also customize computing models, demonstrating strong heterogeneous compatibility and flexible resource management.

Full-Link Real-Time Monitoring   The platform provides granular real-time monitoring from clusters and nodes down to individual instances. Testing showed that long-cycle training tasks can be dynamically tracked with instant and accurate indicator updates. When resource utilization exceeds thresholds, the alert system triggers notifications immediately. During task execution, data remains fully visible, ensuring stable operation and comprehensive fault diagnosis for large-scale computing workloads.

Leading the Industry, Continuously Expanding Native AI Cloud Capabilities

With mature and stable computing service capabilities, SenseCore has already supported numerous large model developers and industry clients, widely covering scenarios such as AIGC, embodied AI, AI4S, and industrial intelligence. It plays a critical role in enabling the deployment of large models in real-world applications.

SenseCore remains firmly in the leading tier of the market. According to Frost & Sullivan’s China Full-Stack AI Cloud Service Market Report, H1 2025, SenseTime ranked first among Native AI Cloud providers in overall market share. IDC’s China Large Model Platform Market Share, 2024 report showed that SenseCore held a 12.2% market share, ranking third overall and solidly within the industry’s top tier.

Looking Ahead

SenseCore will continue to strengthen its “Native AI Cloud” capabilities, upgrade its computing foundation, and accelerate the scaled deployment of large models in real business scenarios.