On May 20, at the 2026 Alibaba Cloud Summit, Alibaba Cloud officially launched a new AI service platform called "Qwen Cloud," designed for the Agentic era. Positioned as "the full-stack intelligent infrastructure for AI Agents," this product marks a shift in the cloud computing paradigm from a compute-centric model to an agent-centric one, completely redefining the service pipeline in the large model era.
The core highlights of "Qwen Cloud" lie in achieving full "Skillization" and "CLIization" of model services. By encapsulating complex processes such as model selection, resource invocation, authentication configuration, and usage query into standardized tool interfaces, AI Agents can autonomously access platform capabilities with just commands, without requiring manual coding for integration. Currently, the platform has aggregated more than 150 model series, including Tongyi Qianwen Qwen3.7-Max, Zhipu GLM, Moonshot Kimi, DeepSeek, and others, totaling over 480 mainstream models. Among them, the flagship model Qwen3.7-Max ranks first among domestic models in the blind test overall ranking, demonstrating exceptional alignment with task objectives.
On the underlying support side, Alibaba Cloud also launched the Panjiu server based on the new self-developed AI chip "Zhenwu M890," which offers three times the performance of the previous generation, with point-to-point latency reduced to within 150ns. Combined with the newly upgraded "Agentic Cloud" infrastructure, cloud products have completed lightweight sandbox environment adaptation for Agents, capable of handling special workloads such as high-frequency, short-life cycle, and sudden high concurrency during agent operations.
In addition, the platform introduced an innovative "Token Plan" subscription model, aiming to reduce the cost of frequent AI programming and agent tools through a more flexible billing method. This move signifies that cloud services are about to fully enter the "agent-native" stage, accelerating the transition of AI applications from single-generation to complex task automation.
