'Reliable AI' Evaluation Framework for Agent Era Launched at AICC 2026

By AI123

On September 22, at the 2026 AI Computing Conference (AICC 2026), the Fifth Institute of Electronics of MIIT, together with Beijing Academy of Artificial Intelligence (BAAI), Inspur Information, Alibaba Cloud and more than 20 industry chain partners, officially released a "Reliable AI" evaluation system designed for the Agent era.

The new framework addresses the limitations of traditional AI evaluation that focuses on individual models, chips or partial performance. It covers chips, systems and interconnect, frameworks and software stacks, models, data and applications, providing full-chain assessment from reliability, security and stability perspectives. The system includes scenario construction, risk assessment and capability optimization, offering verifiable evaluation references for industry selection, technology evolution and application deployment.

In addition, IDC and Inspur Information released the "2026 China Artificial Intelligence Computing Development Assessment Report" at the conference, noting that agents are driving a surge in token consumption and projecting China's intelligent computing power to reach 2576.5 EFLOPS in 2026, up 87.9% year-on-year.

Explore topicsAgentAIAICC2026Policy

Source

View all