At the 2025 World Artificial Intelligence Conference and the High-Level Meeting on Global Governance of Artificial Intelligence (WAIC), independent distributed cloud computing service provider PPIO officially launched China's first Agentic AI infrastructure platform, marking a new development stage in agent application development.
Two Product Versions Cover All Scenario Needs
PPIO Co-founder and CEO Yao Xin introduced at the WAIC event that the AI agent platform is divided into two product lines: the general version and the enterprise version. The general version is supported by a distributed GPU cloud foundation, providing individual developers and small and medium-sized enterprises with an intelligent agent development environment that combines performance, efficiency, and cost advantages. The enterprise version targets mid-to-large enterprises, building a complete closed-loop system from intelligent decision-making to execution through five differentiated advantages: long-term memory, controllable processes, full-domain operations, business integration, and multimodal collaboration.
The First Agent Sandbox Compatible with E2B Interface Shockingly Launched
PPIO's Agent sandbox is a cloud-based runtime environment specifically designed for agents to perform tasks. Built on Firecracker MicroVM, it has three core features: strong security isolation to ensure complete separation of different agent sandbox environments, preventing data leakage and resource conflicts; sub-200ms millisecond-level rapid startup, significantly improving development smoothness; and the ability to create high concurrency of thousands of sandbox instances, meeting business high-concurrency needs.
This sandbox supports dynamic calls to various tools such as Browser use, Computer use, MCP, RAG, and Search. It has already integrated well-known open-source projects like Camel AI, OpenManus, and Dify, giving agents safe, reliable, and efficient execution capabilities.
Comprehensive Model Services Support the "Brain" of Agents
To meet the decision-making needs of agents, PPIO model services support quick access to hundreds of mainstream open-source and custom AI models. In terms of planning capabilities, DeepSeek R1, Qwen3, and MiniMax M1 have been launched as mainstream domestic inference models; in terms of memory needs, DeepSeek's context window has been upgraded to 160K, meeting the demand for multi-turn long conversations; in terms of execution capabilities, Kimi-K2 and Qwen3Coder code models have been supported, achieving powerful self-programming and tool calling abilities.
The platform also provides support for multimodal models such as text, voice, video, and images. Through its self-developed inference acceleration engine, it deeply optimizes key technologies such as PD separation, FlashMLA, and DeepEP, achieving a dynamic balance between inference efficiency and resource usage.
Smart Education Scenarios Are Initially Implemented
The enterprise version of PPIO's agent platform has already validated its value in real scenarios. Based on this platform, PPIO has developed a comprehensive smart education solution for an international school, covering all scenarios including management, teaching, learning, and research, running through the entire process from pre-class preparation, in-class assignments, learning analysis, psychological monitoring, to home-school interaction, suitable for all educational stages both within and outside the school.
Pushing Agents into a New Stage of Value Creation
Yao Xin stated, "From AI Infra to Agentic AI Infra, PPIO's AI agent platform not only provides comprehensive support for agents and enterprises from underlying computing power to upper-level applications, reducing technical barriers and development costs, but also builds a solid foundation for the large-scale deployment of agents through secure and efficient sandbox environments, diverse model resources, and cost-effective computing power services, pushing agents from technological exploration to a new stage of value creation."
With the launch of this platform, agent application development will gain stronger infrastructure support, and it is expected to accelerate the large-scale application and commercial value realization of agent technology across industries.