Technology newsroom
Microsoft Launches Hybrid Intelligence in Windows 11! 🧠 Seamlessly Merging On-Device and Cloud AI Power
Microsoft introduces Hybrid Intelligence in Windows 11, combining on-device and cloud AI processing power through its core technology, Microsoft Execution Containers (MXC), for maximum efficiency and security. It is specifically designed for modern PCs.
📌 Key Highlights:
- Microsoft introduces a new Hybrid Intelligence feature in Windows 11 to connect on-device AI models with cloud-based models.
- At its core is Microsoft Execution Containers (MXC), a Sandbox technology that allows AI agents to operate securely and with restricted file access.
- Supports various AI models and is designed for modern PCs with NPUs or powerful graphics cards, such as Copilot+ PCs and the RTX Spark chip series.
Microsoft has announced the launch of an exciting new feature for Windows 11 called Hybrid Intelligence. This concept aims to seamlessly and more efficiently integrate the functionality of Artificial Intelligence (AI) models processed locally on our computers (on-device) with large AI models operating in the cloud.
⚙️ Underlying Technology
The core of this feature is Microsoft Execution Containers (MXC), a technology designed to run AI agents within a restricted and controlled environment (Sandbox or Container) to maximize security. It allows specifying which files or resources an agent can access. MXC technology is highly flexible, compatible with various container types such as WSLc, ProcessContainer, Windows Sandbox, and also supports cross-platform operation on other operating systems like Linux (via bubblewrap, LXC) and macOS (via Seatbelt).
🤝 Partners and Supported Models
Microsoft has collaborated with numerous AI industry partners, resulting in many agents already capable of running on MXC, including OpenAI Codex, GitHub Copilot, Replit, LM Studio, and NVIDIA OpenShell. Additionally, there are coding models that can run locally on high-spec machines, such as MAI-Code-1.1-Flash, or other options like NVIDIA Nemotron 70B and DeepSeek V4 Flash.
☁️ Cloud Connectivity
In cases where on-device models lack sufficient capability, Hybrid Intelligence can seamlessly switch to cloud-based models via GitHub HydraFusion. This acts as an intermediary, automatically managing and routing to the most suitable model based on predefined conditions. This feature will be integrated into GitHub Copilot and Visual Studio Code by October, providing developers with convenient access.
💬 How do you think the Hybrid Intelligence feature will transform our workflow on Windows 11 and empower developers with more powerful tools?