As the hype around large models gradually subsides, the real competition for industrial intelligence is just beginning. On October 28, Qihoo 360 Group officially launched its enterprise-level intelligent agent platform in Beijing, announcing the release of the world's first intelligent agent operating system covering all stages from L2 to L4. At the same time, it upgraded its core engine - the SEAF Intelligent Agent Factory, aiming to provide governments and enterprises with an "easy-to-use, reliable, and practical" one-stop AI implementation solution.

From "0 to 1" to "1 to 10": 360 Defines the Evolution Path of Intelligent Agents

At the launch event, Yin Yuhui, Senior Vice President of Qihoo 360 Group and CEO of the Digitalization Group, reviewed the evolution path of intelligent agent technology: In August this year, Qihoo 360 first introduced the world's first L4-level enterprise intelligent agent factory SEAF (Self-Evolving Agent Factory) at the ISC.AI 2025 conference, achieving autonomous perception, decision-making, and execution of intelligent agents. Now, by building a complete development-management-operation platform, Qihoo 360 is driving intelligent agents from single-point breakthroughs to large-scale replication, completing a key transition from "technical verification" to "industrial application".

Addressing Three Major Pain Points for Government and Enterprises: Not Usable, Not Easy to Use, Not Trustworthy

Chu Weiming, General Manager of the Product Planning Center of Qihoo 360, pointed out that government and enterprise organizations generally face three obstacles when introducing intelligent agents: insufficient capabilities leading to "not usable", awkward interactions causing "not easy to use", and lack of security and compliance resulting in "not trustworthy". The SEAF Intelligent Agent Factory was specifically designed to solve these problems.

The platform supports private deployment and adaptation to the information innovation environment, ensuring data stays within the domain and models are auditable; it also provides visual orchestration, multi-agent collaboration, and continuous learning and optimization capabilities, enabling non-technical departments to quickly build business intelligent agents. Whether it's emergency scheduling in urban governance, equipment inspection in the aviation field, or auxiliary diagnosis in medical scenarios, or personalized teaching in the education industry, SEAF has already formed mature cases with a "perception-decision-execution" loop.

Full Lifecycle Management, Building an Enterprise AI "Operating System"

The newly released intelligent agent platform covers the full lifecycle of intelligent agents:

  • In the development phase: It provides a low-code toolchain and industry knowledge base, supporting the rapid construction of intelligent agents from L2 (task-oriented), L3 (planning-oriented) to L4 (self-evolving).
  • In the management phase: It realizes version control, permission allocation, performance monitoring, and security auditing.
  • In the operation phase: It supports A/B testing, user feedback loops, and automatic iteration, ensuring the continuous evolution of intelligent agents.

Launching the "Thousand Industries" Ecosystem Plan, Bridging the Last Mile of AI Implementation

To accelerate industrial penetration, Qihoo 360 simultaneously announced the launch of the "Thousand Industries Industry Ecosystem Cooperation Plan," uniting software and hardware manufacturers, system integrators, and industry ISVs to jointly create standardized intelligent agent solutions for sectors such as government, finance, manufacturing, and energy. Li Fangxiang, Chief Solution Architect of Qihoo 360's Digitalization Group, stated: "We do not develop general large models, but only industry intelligent agents that can solve real problems."