Hello Windows Insiders, Today, we’re releasing new Windows 11 Insider Preview builds to the Beta and Experimental channels! Coming straight off the back of our Windows + Surface event, we have some exciting new features entering the Windows Insider Program. New builds this week Select your channel below to view its release notes:

• Beta: Build 26220.9606
• Experimental: Build 26340.9616

For those on other specific Windows build versions, here are today’s new builds and release notes:

• Beta (26H1): Build 28020.3172
• Experimental (26H1): Build 28120.3181
• Experimental (Future Platforms): Build 29683.1000

Notable new features:

[Windows Search]

Release channel: Experimental, Experimental (26H1)

• New Windows Search experience with inline actions: We're beginning to roll out a new Windows Search experience built on WinUI 3 that is faster, more efficient, and features a streamlined design that makes results easier to scan. We're also introducing inline actions, allowing you to quickly complete tasks directly. Turn on dark mode or Bluetooth, adjust your screen brightness, mute your device, or arrange windows from Windows Search.
• We're improving how Windows Search understands what you're looking for, including better matching for typos and synonyms. We'll continue fine-tuning these improvements over the coming weeks. For more information detailing the new changes, check out the dedicated Windows Insider Blog.

[caption id="attachment_179190" align="aligncenter" width="2560"] Screenshot of the new Windows Search experience[/caption]

[Battery Status widget]

Release channel: Experimental

• As part of our continued commitment to making Widgets more useful for everyday moments, we're introducing the Battery Status widget for Insiders on newer builds. This widget provides a unified view of battery levels for your PC and peripheral devices, helping you quickly assess power status without hunting through multiple UI locations. Add it by going to Widgets board > Add widgets > Battery Status.

[caption id="attachment_179221" align="aligncenter" width="604"] UI showing the new Battery Status widget, providing a unified view of battery levels across your PC and connected devices.[/caption] Thanks, Stephen and the Windows Insider Program team

For more than a decade, Surface has been shaped by the people pushing technology further. Creators rendering complex scenes. Developers building what doesn't exist yet. Designers moving between ideas, models, code, and content. Makers whose work demands more from their devices because they're constantly pushing boundaries. Over time, we've learned something important: the people who choose Surface aren't just using technology for tasks; they're shaping what's next. Today, we're opening pre-orders for two new devices built for that reality: Surface Laptop Ultra and Surface RTX Spark Dev Box. These new Surface devices bring to life the broader Windows vision for hybrid intelligence we’re sharing today, combining the power of the PC and the cloud to help people do more with AI.

Surface Laptop Ultra: Made for world makers

Surface Laptop Ultra started with a clear ambition: How can we better serve the needs of creators, developers, and AI builders? The answer wasn’t simply a faster laptop. It meant fundamentally rethinking what a high-performance Windows PC could be. It started with a deep partnership to optimize Windows to take full advantage of NVIDIA’s RTX Spark™ Superchip. We then designed an entirely new device around the platform from the ground up. Mechanical, electrical, thermal, acoustic, material engineers, and industrial designers worked together from the beginning to fully realize the potential of the RTX Spark. The result is our most powerful Surface Laptop ever, built for demanding creative, technical, and AI workloads.

One superchip. One complete system.

Designed around NVIDIA RTX Spark, Surface Laptop Ultra unites an NVIDIA Blackwell RTX GPU with up to 6,144 cores, an NVIDIA Grace CPU with up to 20 cores, and up to 128 GB of unified memory.¹ Run AI models exceeding 120B parameters locally, with up to 1 petaflop of AI performance.² Move between creative work, development, and immersive gaming without missing a beat. Keep more of your work on-device, with less reliance on the cloud and access to the CUDATM ecosystem and full Windows AI stack.³

Power is only the beginning

Power only matters if it lasts. That's why we designed a new thermal architecture that delivers up to 2.5x more thermal power capacity than our current Surface Laptops. In a chassis less than 18 mm thin and under 4.5 lb⁴, Surface Laptop Ultra was built for sustained performance. The 15-inch PixelSense™ Ultra touchscreen is the brightest, sharpest, and most vivid display we've ever put on a Surface Laptop.⁵ It’s designed to empower professionals to review footage frame by frame, annotate designs, explore datasets, and interact directly with their work. It delivers the brightest peak HDR of any laptop⁶, up to 25% brighter peak HDR than MacBook Pro M5⁷, and the color accuracy creators trust. The touchpad is 30% larger, with advanced haptics for precise control across demanding creative workflows.⁸ Storage is user-removable, so you can upgrade, replace, or secure your own storage.⁹ Surface Laptop Ultra is available in Platinum and Nightfall, so it looks as sleek and sophisticated as it runs.¹⁰ It also features our most complete port selection yet, led by something entirely new: Magnetic Connect, the world’s first built-in magnetic USB-C® charging on a laptop. The included charging cable attaches magnetically to the right-side USB-C port and releases easily. The port remains fully functional for video, data, and other USB-C devices. Other ports include: · Two additional USB-C ports · HDMI · USB-A · An SD card reader · A 3.5 mm headphone jack Drive up to three 4K displays¹¹. Import footage without adapters. One device. Every connection you need.

Built for real work. Always ready to play.

Creative professionals get a display they can trust, graphics acceleration when projects get demanding, and up to 128 GB of unified memory to provide headroom for multitasking across demanding creative and local AI workflows. Developers can use local models with GitHub Copilot to bring agentic coding on device. They can prototype agents, review code, test larger models, and keep development tools running side by side. When the work is done, the same RTX Spark architecture that accelerates creative and AI workloads also powers a great gaming experience. Ray tracing, DLSS and NVIDIA Reflex in supported titles, together with the deep and growing catalog of games on Windows on Arm, mean the machine you rely on during the day can keep delivering when it’s time to play.

Surface RTX Spark Dev Box

Surface Laptop Ultra moves with you. Surface RTX Spark Dev Box gives developers, AI engineers, and technical teams powerful AI compute at their desks. Also built around NVIDIA RTX Spark Superchip, this developer-first powerhouse brings massive local AI compute to a compact small form factor, giving developers more control over where workloads run and how resources are used. When experimentation happens locally, teams can iterate faster, reduce cloud dependency, and reserve cloud resources for the workloads that truly need them. Dev Box offers developers:

• Up to 1 petaflop of AI performance and 128 GB of unified memory on your desk
• Support for AI models exceeding 120B parameters with large context windows²
• Local inference, evaluation, and experimentation at interactive speeds
• The same powerful NVIDIA RTX Spark platform that powers Surface Laptop Ultra

Ready to go out of the box The goal is simple: less time configuring your environment, more time building. As part of the Project Zenith family of devices, Surface RTX Spark Dev Box comes ready with the tools, runtimes, and environments developers use most, including Visual Studio Code, Git, GitHub CLI, GitHub Copilot, WSL, Python, Node and more.

Run AI agents where they work best

Give developers the freedom to move faster with agents, while keeping access, compute, and cost under control. Windows is the OS for agents, and Surface is the hardware for Windows. Microsoft Execution Containers provide containment for agents, while upcoming Windows capabilities will use Microsoft Entra to distinguish agent activity from user activity and extend Microsoft Agent 365 controls to local agents on-device. With NVIDIA RTX Spark and up to 128 GB of unified memory, Surface Laptop Ultra and Surface RTX Spark Dev Box can run demanding models and agent workflows locally, including GitHub Copilot tasks that investigate builds and propose fixes, while Microsoft Dev Box adds cloud capacity when needed.¹ The result is governed, scalable AI development without spending a cloud token on every step. Commercial configurations of Surface Laptop Ultra are available for pre-order today.

Break up with your MacBook

If you've outgrown the compromises of a traditional professional laptop, we'll help make the switch. Level up to Surface Laptop Ultra and get up to $1,000 cash back when you trade in an eligible MacBook Pro.¹² We're confident that once today's creators and developers experience everything Surface Laptop Ultra has to offer, they won't look back. Learn more about the offer here.

Reserve yours today

Starting today, pre-order Surface Laptop Ultra starting at $2,599 (MSRP) from Microsoft.com or select retailers worldwide, with availability beginning October 16. At Microsoft Store, explore the full lineup of configurations and accessories, with free shipping, easy returns, flexible payment options, and Microsoft support. If you’re looking for a commercial device, explore Surface Laptop Ultra for Business. Surface RTX Spark Dev Box is available for pre-order for $5,999 (MSRP), exclusively on Microsoft.com in the U.S. and will begin shipping to customers this November.

---

¹ Memory varies by configuration. 128GB total system memory in a unified memory architecture dynamically shared between CPU and GPU. Maximum amount addressable by the GPU depends on system configuration and workload and is less than the total. ² Based on theoretical FP4 performance of 1 petaflop using the sparsity feature. See: NVIDIA Corporation - NVIDIA and Microsoft Reinvent Windows PCs for the Age of Personal AI. ³ CUDA™ is a registered trademark of NVIDIA Corporation. ⁴ Might vary depending on manufacturing processes. ⁵ Surface display has rounded corners within a standard rectangle. When measured as a standard rectangular shape the screen is 15” diagonally (actual viewable area is less). Touchscreen is optimized for touch interaction only. Pen input and compatibility with Surface Pen, Surface Slim Pen, or similar accessories is not supported. ⁶ Peak HDR brightness of 2,000 nits, measured over a 10 percent window with HDR content. Comparison against laptops commercially available as of October 2026. Sustained and SDR brightness are lower. Actual brightness varies with content, settings, and conditions. ⁷ Based on peak HDR brightness comparing Surface Laptop Ultra to MacBook Pro M5 Pro published specifications as of October 1st, 2026. ⁸ Compared to Surface Laptop 8th Edition ⁹ Replacement components and service options vary by market and over time. Repairs should follow the Microsoft Service Guide. Device damage caused during repair may not be covered by warranty unless required by law. ¹⁰ Colors available on selected models only. Available colors, sizes, finishes, and processors may vary by store, market, and configuration. ¹¹ External monitor support requires compatible displays, cables, and accessories. See Surface tested peripherals page for compatibility information. ¹² Offer valid with purchases of qualifying new Surface Laptop Ultra device between October 7, 2026 - November 23, 2026 after trade-in of a qualifying device, where customer submits the online trade-in claim for their used device, plus proof of purchase for new device, during the Promotion Period. Available only through the Microsoft Store online in the United States and Canada. Amount of additional cash back varies depending on device purchased and value of trade in; see FAQ for more details. To be eligible for the cash back, customer must mail in the qualifying trade-in device within 15 days of completing the online trade-in process. To be eligible for trade-in, (i) customer must own the qualifying device, (ii) the device must power on, (iii) the battery must hold a charge (the device must not be required to be plugged in to operate), and (iv) the device must be in fully functional, working condition and not have broken or missing components, a cracked display or screen, liquid damage, modification(s) or the device warranty seal broken. The device cannot be password protected and must include its original chargers, accessories, and hard drive. Trade-in value depends on device type and condition after inspection by Teladvance. All trade-ins and reimbursements are subject to Teladvance's sole discretion and approval. Devices determined to not meet eligibility criteria may be shipped back to the customer or recycled. Recycled devices are not eligible for cash back. All trade-ins are final. See Microsoft Store Trade-in Program for steps and requirements. Not valid on prior orders or purchases; cannot be transferred or otherwise redeemed for cash or gift cards, or credit towards other devices or products. May not be combined with other offers. Void where prohibited or restricted by law. Microsoft reserves the right to modify or discontinue offers at any time. Limit of 1 offer per customer.

Agents are unlocking enormous productivity gains for customers, but their ability to work across files, networks, and applications can introduce new security risks. This can leave customers feeling like they only have two choices: give agents unrestricted access and hope nothing goes wrong or block them and lose the productivity benefits they provide. Neither option is acceptable. That’s why we’re building Windows platform capabilities to help run and manage agents more securely, starting with:

• Containment: limiting what an agent can access and do.
• Identity: distinguishing an agent’s activity from a person.
• Manageability: giving organizations the tools to govern access and monitor agent activity.

Microsoft Execution Containers (MXC), now generally available, provides the containment layer. Developers and IT administrators define the resources, like files and network destinations an agent can use and MXC uses the appropriate container to enforce those policies at runtime.   Windows will also soon enable Microsoft Entra to help distinguish agent activity from user activity, ensuring users can remain productive even when agent access needs restrictions and extend Microsoft Agent 365 controls to local agents on-device, enabling IT teams to manage MXC containers, apply policies, and monitor agent activity.

Why agents need a managed execution boundary 

An agent cannot be its own security authority. It must run within a boundary defined by the developer or organization and enforced independently of the agent itself. Consider a coding agent asked to update a website. The agent needs read and write access to the website repository and access to the development tools required to build and test the change. It may need to read production server configuration to understand how the application is deployed, but it should not be able to modify that configuration. Without a managed execution boundary, the agent may decide that changing the server configuration is the fastest way to complete the task and potentially break the production site. The action can be reasonable from the agent's perspective and still exceed the authority the developer intended to grant. Containment creates that managed boundary: the agent can read and write the repository, read the server configuration, but is not authorized to access anything else unless access has been granted. If the agent attempts to modify the server configuration, the containment environment is designed to prevent the operation regardless of what the model, generated code, plugin, or tool decides to do.

What is Microsoft Execution Containers (MXC)? 

Microsoft Execution Containers (MXC) is a policy-driven execution layer for untrusted code or dynamically generated workloads. In agentic scenarios, developers can use MXC to contain model-generated output, plugins, tools, agent harness, or the entire agent. This limits what the agent workload can access by containing the scope of the impact if something goes wrong.   Developers declare the resources a workload requires, such as files and network destinations, and MXC enforces the resulting boundary with the appropriate container. The policy remains outside the agent workload’s control, so the agent or generated code cannot grant itself additional access. MXC also separates workload requirements from platform-specific containment details, greatly simplifying the developer experience. Developers integrate with a unified JSON configuration schema and multi-language SDK, while MXC maps the requested controls to the selected backends on Windows, macOS, or Linux. Developers can apply the same containment model wherever an agent needs to run – from a local device to the cloud. With Windows 365 support for MXC now generally available, developers can run agents alongside their existing work on Cloud PCs, using the right isolation model to help keep agent execution separate and secure.

Choose the containment level that fits the workload 

Different workloads require different levels of isolation. A coding agent working in a repository may prioritize low latency and responsiveness, while an agent processing sensitive data or running untrusted code may require stronger isolation. MXC provides a spectrum of containment options so developers and organizations can select the right level of isolation for each workload. Only Windows supports a session container, which runs an agent on the user’s device in a separate, OS-isolated session with its own local agent identity and isolated desktop, clipboard, UI, and input boundaries.

Backend	Availability	Best suited for	Important characteristics
Process container	Windows 11, macOS, and Linux	Lightweight containment for responsive workloads, including model-generated code and tool execution.	Uses the platform-appropriate process sandbox, including AppContainer on Windows, Seatbelt on macOS, and Bubblewrap on Linux.
Session container	Windows 11 only	Long-running agents and automation that need a desktop or stronger separation from the interactive user.	Runs under a distinct Windows account and session, separating the agent's desktop, clipboard, UI, input, and active session from the user.
WSL container (WSLc)	Windows 11 only	Linux-first agent toolchains and workloads that depend on the Linux package and development ecosystem.	Provides a Linux execution environment through WSL.
MicroVM	Windows 11 and Linux, experimental	Higher-risk workloads that benefit from a hardware-backed virtualized boundary.	Provides hardware-enforced isolation and full Linux workload compatibility.

Each containment backend has distinct security properties and workloads should be evaluated for fit with containment backend.

How MXC policy defines workload boundaries 

MXC helps users and organizations delegate more work to agents while limiting them to the resources required for each task. Instead of giving an agent the full authority of the signed-in user, developers and IT can define an OS-enforced boundary around the agent’s workload.  For the website scenario, an MXC policy could give a coding agent read and write access to its local source-code repository and access to required development tools such as Git, while preventing access to personal locations such as the user’s Documents folder. The policy could also block inbound and outbound network connections and access to the interactive desktop.  

Policy area	What it controls
Containment	The isolation environment in which the workload runs, such as a process or session container.
Process	The command, arguments, working directory, environment, and other settings used to start the workload.
File system	Locations the workload can modify, locations it can read without changing, and locations it cannot access.
Network	Inbound and outbound connectivity, including whether the workload can connect to services through the host’s loopback interface.
User interface	Whether the workload can access or interact with the desktop and related UI resources.

Adding MXC support is straightforward for agent developers – you can easily use your favorite coding agent to integrate the MXC SDK and draft an initial workload policy, then review, test, and refine the resulting controls.

How workload and organizational policy work together 

Agent developers declare the resources their agentic workloads may need. Organizations can also apply additional constraints through management policy, like with Microsoft Intune management policy. This allows the same agent to operate within different enterprise boundaries without requiring the agent developer to encode the organization's security posture into the application. Containment becomes more valuable when organizations can apply it consistently. Intune policy will soon be available to manage MXC process containers used by MXC-integrated agents on Windows 11. These policies will offer IT administrators control over how Windows evaluates container creation requests made by agents, and the resource boundaries that are enforced by those containers. Agent developers should design their workloads to operate within boundaries that may be more restrictive than the default configuration. If organizational policy blocks a resource, the agent should explain that the task could not be completed within the available permissions, request an appropriate user or administrator action when supported, or choose a safe alternative. It should not silently fail.

Observe and refine policy before enforcing it 

Writing a least privilege policy can be difficult when you do not yet know every resource an agent workload requires. Only on Windows, MXC process containers can produce an agent activity report showing which resources a workload attempted to use to help craft a least-privilege policy. MXC supports three operating modes:  

Mode	Ungranted access	Activity Report	Intended use
Enforcement	Blocked	No	Run the workload with its production policy
Learning	Blocked and recorded	Yes	Diagnose failures and verify that the policy grants only required access
Permissive	Allowed and recorded	Yes	Observe agent activity without enforcing policy

In Enforcement mode, MXC applies the policy without activity report. Granted operations proceed, and operations outside the boundary are restricted. In Learning mode, MXC continues to enforce the configured boundary. An operation that has not been granted is blocked and recorded in a JSON activity report. This lets developers or IT reproduce containment failures and understand which resources the workload attempted to access. In Permissive mode, MXC records access that the policy would have denied but allows the operation to continue. This is useful during policy authoring because the workload can be completed while developers or IT collect evidence about the resources it uses. Permissive mode does not bypass other applicable operating system or organizational restrictions. When you bring an agent into an MXC container for the first time, the policy may block capabilities that the workload legitimately needs. These denials reveal where the boundary needs adjustment, helping you grant the required access without unnecessarily expanding the agent’s reach.

Agent identity & attribution 

Containment answers what an agent may do. Identity helps determine which agent performed an action. Coming soon, Windows will allow Microsoft Entra to distinguish agent activity from user activity in Microsoft Agent 365. This separation will allow security teams to evaluate an agent's behavior and risk independently from the person using the device. If an agent becomes compromised or violates policy, security controls will target the agent's access to protected resources without blocking the employee’s access. As organizations deploy more agents, one misbehaving agent does not have to interrupt the employee’s access or productivity. Agent-level attribution also improves investigation and governance. Administrators will be able to use Agent 365 to understand which agents are running, associate activity with a specific agent, investigate risky behavior, and apply policy to individual agents or groups of agents instead of relying only on user-level or device-level context.

Agents using MXC today 

NVIDIA has integrated OpenShell into MXC, complementing MXC with policy controls for agent access to files and inference services; advanced network controls; credential management, and for enterprises, OCSF auditing. Leading agents and agent frameworks already support MXC, including GitHub Copilot, OpenClaw, OpenAI Codex, Replit, LM Studio, and Unsloth AI. Many others will be releasing support for MXC: Anthropic Claude Code, Box, Egnyte, Heidi Health, Hermes Agent by Nous Research, Manus, Perplexity, Raycast, and Simular amongst others. For people using AI every day, these safeguards are part of the agent experience itself, giving users a more secure way to work with these tools. GitHub Copilot, Codex, and Replit agent scenarios demonstrate how MXC supports a demanding developer workflow. A coding agent needs access to the project repository, development tools, and commands required to complete a task, but it should not automatically gain access to unrelated files or network destinations. MXC preserves the capabilities the agent needs while enforcing clear boundaries around the rest of the system. Learn more about how GitHub Copilot uses MXC for sandboxed execution in this blog post. 

Get started and share feedback 

With containment, identity, and manageability built into the platform, Windows helps agents run more securely while giving developers and IT control over what they can access and do. MXC continues to improve as we expand containment backends, policy authoring, manageability, observability, and more supported scenarios in the containers. You can get started today with the MXC SDK, configuration schema, documentation, and samples in the MXC repository. If you encounter unexpected behavior or have ideas for improving the experience, open an issue on GitHub and help shape what comes next.

Agents are creating a foundational shift in how people get things done, unlocking new levels of productivity and capability on the PC. As agents become part of everyday work, they also raise the bar for compute, security, and control.   That’s why we’re building Windows as the home for hybrid intelligence: a platform where agents can run locally when it makes sense, reach the cloud when they need to, and operate with the security and manageability organizations expect.  Hybrid intelligence expands what’s possible. Windows makes it practical, secure, and ready for broad adoption. And powerful new silicon, devices and models unlock local compute. Together, they define the next chapter of the PC: 

• Windows is the most secure platform for agents. With containment, identity, and manageability built into the platform, Windows helps you control what agents can access, know which agent acted, and govern them at scale. Microsoft Execution Containers (MXC), now generally available for Windows and integration with Microsoft Agent 365, are key steps in making that real.   

• Hybrid intelligence brings frontier capability to Windows locally. With intelligent routing, powerful local models, and a performant runtime, Windows lets work run in the right place for each task — local or cloud. We’re bringing frontier-class capabilities to local PCs and enabling services like GitHub to route intelligently on Windows. 

• Copilot gets supercharged with Windows hybrid intelligence. With your permission, Copilot will be able to use local context from your PC, take action on your behalf, and tap into local models, bringing you more capability while helping your tokens go further. 

• A new generation of PCs built for these possibilities. From always-on mini desktop PCs to deskside supercomputers powered by DGX Station for Windows, we are bringing a broad range of devices to unlock hybrid intelligence. Surface Laptop Ultra, powered by NVIDIA RTX Spark™, is available for pre-order today, alongside new devices from ASUS, Dell, HP, Lenovo, and MSI. 

• Windows gets more capable across everyday tasks including new ways to take action directly from Windows Search on the taskbar and enhancements in gaming. 

Windows is the most secure platform for agents

https://www.youtube.com/watch?v=MzhMPCeFc5w Agents don’t work like traditional apps. They can run around the clock, use tools, write code, access files, and act across systems, often without someone watching every step. That requires a new security model built for how agents actually operate. On Windows, that model comes together through containment, identity, and manageability. Containment controls what an agent can access. Identity makes clear which agent acted, so IT can distinguish agent activity from the person using the device. And manageability connects that activity into tools like Agent 365 and Intune, so organizations can set policy, monitor risk, and govern agents across the organization¹.  Together, these primitives allow organizations to secure, observe, and govern agents with confidence.

Expanding the ecosystem of agents on Microsoft Execution Containers, now generally available

Traditional sandbox solutions weren’t designed for agents, where containment needs can change with every prompt and tool call. We built Microsoft Execution Containers (MXC) to meet that need. Today, Microsoft Execution Containers (MXC), becomes generally available on Windows 11, enabling organizations to define which files and networks agents can access, with those policies enforced at runtime. For individuals using AI on their personal PC², these safeguards are part of the agent experience itself, giving users a more secure way to work with these tools. While MXC works across operating systems, its deep integration with Windows gives organizations more control of their data and a broader range of containment options, from process and session isolation to WSLc, virtual machines, and Windows 365 for Agents. Leading agents already support MXC, including Codex from OpenAI, GitHub Copilot, OpenClaw, Replit, LM Studio, OpenShell from NVIDIA, and Unsloth AI. Many others will be releasing support for MXC: Anthropic Claude Code, Box, Egnyte, Heidi Health, Hermes Agent by Nous Research, Manus, Perplexity, Raycast, and Simular amongst others. We’re also excited to share that Muse for Windows, a personal AI agent from Meta, is coming soon as a native app on Windows with MXC integration, bringing another leading AI experience to Windows and providing more choice in the agents people can use on their Windows PC. To learn more about how developers can integrate their agents with MXC, visit: Microsoft Execution Containers: Policy-driven containment for AI agents. 

Bringing hybrid intelligence to Windows

As AI models grow in size and capability, customers’ needs are outpacing what their cloud budgets can support. They want to make every AI token count without giving up frontier capabilities. Hybrid intelligence brings local and cloud AI together, helping customers stretch their budgets while preserving access to the intelligence they need.   To bring this to life requires powerful local models, intelligent orchestration, a performant runtime, and capable silicon, all working together to create a meaningful experience.

Frontier coding intelligence, running locally on your PC

At Build, we introduced MAI Code 1.1 Flash, a 137 billion total and 6.8 billion active parameters, designed for real-world coding workloads. Today, we’re bringing that intelligence to your device, using 3-bit precision to reduce the model size by nearly 80% while preserving coding quality and supporting a 256K context window locally.  In addition to MAI Code 1.1 Flash, we are bringing other powerful models like an upcoming Nemotron model from NVIDIA – with over 70 billion parameter and quantized in 2-bit to use just over 20GB of memory and DeepSeek V4 Flash, a 284B parameter model, and others for optimal speed and intelligence running locally on RTX Spark.  With optimization and tuning, these models bring what was frontier-class capabilities to local PCs while preserving accuracy and performance — using less memory and fewer system resources.  The goal is to keep AI responsive while leaving room for everything else you do on your PC.

Introducing intelligent routing on Windows with GitHub HydraFusion

https://youtu.be/yzBVjfwgWw8 Earlier this year, GitHub launched HydraFusion, which routes each task to the right model in the cloud.   Today, we’re extending that intelligence to Windows, enabling HydraFusion to tap into models running locally on device, making your tokens go further.   Hybrid intelligence on Windows, powered by HydraFusion is coming to the GitHub Copilot app, GitHub Copilot CLI and Visual Studio Code in experimental preview later in October.   To learn more about how we’re working with GitHub to make this possible on Windows, read: Bringing local models and sandboxed tools to Windows and GitHub Copilot. 

Bringing llama.cpp to Windows ML to provide easier access to open-source models

To power this growing ecosystem of highly capable local models, you need a performant runtime. Windows ML is a scalable runtime to deploy models across GPU, NPU, and CPU, and we’re continuously evolving it.   Today, we’re also announcing llama.cpp support in Windows ML, giving developers more open-source model choices, and the ability to experiment quickly with emerging models as they land.  To learn more, visit AI Development on Windows: from PyTorch and llama.cpp to Windows ML. 

Supercharging Copilot on Windows with hybrid intelligence

Recently, we introduced the new Copilot built around three experiences: Home, Code, and Autopilot. Today, we’re announcing that on Copilot+ PCs, Copilot will get supercharged with Windows hybrid intelligence, providing three key local capabilities.   

• Local context. With your permission, Copilot can understand relevant content on your PC, including files and recent activity, bringing local context into the flow of work. 

• Local actions. Copilot can take action on your behalf across Windows, helping complete tasks like organizing files, assessing device diagnostics, troubleshooting issues, coding, and carrying out workflows directly on your device. 

• Local models. Copilot can tap into AI models running on your PC, using local compute for the right tasks while combining it with cloud intelligence when needed. 

Across Home, Code, and Autopilot, hybrid intelligence makes Copilot more useful on Windows.  In Home, Copilot can use local context from your PC to pull in the files you’ve been working on and help turn them into collaboration-ready artifacts.  In Code, hybrid intelligence helps you create native Windows apps from a single prompt, run code more securely with MXC support, and use local models when appropriate so projects can move faster and token costs can be managed.  And hybrid intelligence will make Autopilot, a persistent, proactive and personal agent, even better. Hybrid intelligence brings local context, local models and PC actions into the agent flow, enabling more grounded personal agent experiences that can continue working on your behalf, even while you focus on other things.  This is the next step for Copilot on Windows, powered by hybrid intelligence. Copilot understands your PC, takes action with your permission, and uses local models when it makes sense, with you in control at every step. Copilot features powered by hybrid intelligence are expected to begin rolling out on Copilot+ PCs in the coming months.³

Offering a range of PCs for agentic work

To bring our vision of hybrid intelligence and new experiences to every desk and every home, you need a broad range of silicon and Windows PCs, from mini desktops up to deskside supercomputers.

Mini desktop PCs, with a simplified out-of-box setup for running agents 

https://youtu.be/DsLkQ04iECA People are already using mini desktop PCs as always-on machines for long-running agents. But too often, getting started requires terminal-based setup, complicated configuration, and security trade-offs that weren’t designed for agent workloads.  Today, we’re announcing a simpler way to set up popular agents on mini desktop PCs in just a few clicks, right out of the box. For OpenClaw users, we’re introducing a native Windows gateway that simplifies setup, along with MXC integration to help keep agents securely contained in a sandbox.  This is just the beginning. Over time, we intend to bring these experiences to more Windows devices, guided by customer feedback, so more people can securely run and work with the agents they choose.

Copilot+ PCs, with hybrid intelligence

Copilot+ PCs established a new class of thin, light and powerful Windows PCs designed to deliver advanced AI experiences on device. Across all Copilot+ PCs, over 2 trillion inferences are run locally per month, improving customers’ experience across search, imaging, and video calls, and today, over 40% of laptops being built for business are Copilot+ PCs. With fast performance, great battery life, and built-in AI features, Copilot+ PCs are your PC for everyday tasks, with even more capability on the way with the new Copilot and hybrid intelligence. See the latest Copilot+ PCs from Acer, ASUS, Dell, HP, Lenovo and Microsoft Surface, at: https://www.microsoft.com/en-us/windows/shop-pcs.   Copilot features powered by hybrid intelligence are expected to begin rolling out over the coming months on Copilot+ PCs⁴.

Builder PCs, for a new way to create with cloud and local AI

We’re designing new PCs for the builders – the developers, creators, and makers who are pushing the limits of what they can do with their PC. They bring together high-performance local compute and intelligence so people can build, iterate, and create with the right AI capability for the task. Builder’s PCs require powerful silicon, unified memory, and optimized AI experiences built for demanding local workloads. In June, Microsoft and NVIDIA introduced RTX Spark to help usher in this category. Windows will continue partnering with NVIDIA and other silicon vendors to expand these capabilities across more devices. Today, we’re excited to announce pre-order of new Windows PCs powered by RTX Spark from ASUS, Dell, HP, Lenovo, MSI and Microsoft Surface.

Surface Laptop Ultra and Surface RTX Spark Dev Box, available for pre-order

Surface Laptop Ultra, our most powerful Surface ever, brings the vision of hybrid intelligence to life for creators, developers and AI builders. Designed around the RTX Spark, it combines up to 128 GB of unified memory with the ability to run AI models exceeding 120 billion parameters locally, bringing powerful AI compute to demanding creative and technical workflows. Alongside Surface Laptop Ultra, Surface RTX Spark Dev Box brings powerful local AI compute to the desk, helping developers and AI engineers experiment, evaluate models and iterate locally. Both devices are available for pre-order today, with Surface Laptop Ultra available beginning October 16 and Surface RTX Spark Dev Box shipping to customers in the U.S. November.

An ecosystem of Builder PCs, powered by RTX Spark

With Windows on RTX Spark, we're bringing last year’s frontier capabilities to your laptop. New Windows PCs powered by NVIDIA RTX Spark were designed as an ecosystem to maximize performance in tasks like local inference, image and video generation. We’re proud to support our device partners as they meet the unique needs of our customers. Across Windows PCs with RTX Spark, builders can see significant local AI performance gains compared with an Apple MacBook Pro 16-inch with M5 Pro, including up to⁵:

• 2.1x faster time to first token⁶
• 4.3x faster AI image generation⁷
• 6.2x faster AI video generation⁸

Today you can pre-order the following PCs: 

• ASUS ProArt P16 and ASUS ProArt P14: Combines powerful AI performance, ASUS Lumina Pro OLED displays and big capacity battery life in slim, lightweight 16- and 14-inch designs built for creators on the go.

• Dell XPS 16 Creator Edition: The first XPS laptop powered by NVIDIA RTX Spark, giving creators and developers studio-grade performance, professional display accuracy, and local AI development in a premium, portable design.

• HP OmniBook Ultra 16: Combines an ultra-thin premium design with breakthrough thermal innovation, delivering sustained performance for the most demanding AI workloads.

• Lenovo Yoga 9n 2-in-1: A versatile convertible that combines NVIDIA RTX Spark with a 360-degree design and dual-surface pen input for creators who need high performance and flexibility to work independently or collaborate.

• MSI Prestige N16 Flip AI+: Combines a thin-and-light 2-in-1 design with a 16-inch UHD+ Tandem OLED display, NVIDIA AI acceleration and a 99.9Wh battery for creators, professionals and gamers on the go.

• Surface Laptop Ultra: The most powerful Surface Laptop ever, built around NVIDIA RTX Spark with up to 128GB of unified memory, a 15-inch touchscreen and a new thermal architecture engineered for sustained performance. 

All of these RTX Spark Windows PCs are available for pre-order starting today and begin shipping October 16. The RTX Spark based dev boxes and mini desktops will be coming later this year.

AI supercomputers, with DGX Station for Windows

With Windows on DGX Station we're bringing last quarter’s frontier capabilities to your desktop. At the top of the range, we’re excited to announce that new supercomputers powered by NVIDIA DGX Station will be available on Windows later this year. Utilizing the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, DGX Station for Windows brings data-center-class performance together with enterprise grade security and manageability. This unlocks the power to run frontier-class models locally, from Llama 4 Maverick and Kimi K2.6 to DeepSeek V4 Pro, at more than a trillion parameters⁹. Capabilities that once required renting a cluster, now in a deskside supercomputer. This brings a new level of AI compute to Windows, enabling developers, engineers, scientists and organizations everywhere to run frontier-class models and agentic workloads on devices that were previously limited to the cloud or data centers. The DGX Station for Windows enables 2 key customer scenarios:

• Personal AI Supercomputer: For the most demanding engineering, scientific, and research workloads, DGX Station for Windows enables a generation of powerful Windows applications accelerated by multimodal agents that can model, perceive, and design for the physical world.

• Scalable Windows Token Factory: DGX Station for Windows can also serve as powerful token factory for teams, supporting 32 or more simultaneous agents serving teams of developers, engineers, or researchers, helping reduce cloud token consumption.

By keeping inference local, organizations can keep sensitive data and IP within their own environment and manage their AI Supercomputers and Token Factories using the same trusted tools they use across their Windows fleet today. The result is a shared AI resource on Windows that helps lower operating costs, improve responsiveness, and give teams greater control over how and where their AI workloads run. Paired with MXC, it’s never been easier to build, govern, secure, and trust end-to-end agentic loops on Windows. Today we welcome the Dell Pro Precision with GB300, and HP ZGX Fury AI Station, both coming later this year, alongside more from our other OEM partners, running the Windows platform organizations know and trust.

Windows gets more capable across everyday tasks

Hybrid intelligence opens up an entirely new chapter for Windows, but innovation isn’t only about what’s next. It’s also about making the experiences you rely on every day faster, easier and better, from getting things done to how you play. 

Introducing actions in Windows Search: Go from finding to doing

https://youtu.be/Nj5KL9yDlwE Search has always been about quickly helping you find what you need. Today, we’re introducing a new Windows Search experience that is faster, more efficient, and enables richer experiences, beginning with thousands of new actions that will be available directly from the taskbar.  These new actions help you finish tasks right where you started, without opening another app or navigating through additional screens. We’re starting with familiar tasks where saving a few steps can make a meaningful difference in everyday use.    Here are a few things to try:  

• Turn something on or off. Try phrases such as “switch to dark mode, “turn on Do Not Disturb,” or “enable Airplane mode.” 
• Make a quick adjustment. Try “dim my screen” or “mute.”
• Manage your workspace. Try “minimize all my windows,” or “arrange my windows.”
• Send a message. Try “Text Elisa I’ll be there soon.” 

We will also begin adding the option to integrate the new Copilot with Search on the taskbar, so you can get quick answers inline from Search, and if you want to have a richer conversation, you can jump from Search directly to the full Copilot app. This experience will be opt-in, and will first start rolling out to select markets later this year. New actions will begin rolling out to Windows Insiders in the experimental channel today. Visit to read more: From searching to doing: Building a faster, more streamlined Windows Search . 

Advancing gaming across the Windows PC ecosystem

Gaming is a fundamental part of Windows, and with new Windows PCs powered by RTX Spark, we’re continuing to expand where and how people can play the games they love.  Over the past several years, we’ve made significant investments to improve the gaming experience on Arm-based Windows PCs with expanded availability of Prism, support from leading anti-cheat providers, and XBOX PC app compatibility.   RTX Spark™ now raises those investments to new levels for gaming performance and graphics capabilities, showcased by Gears of War: E-Day, with support for features like DirectX raytracing, variable rate shading, and DirectStorage. Gears will also be the first to offer expanded support for Advanced Shader Delivery, cutting first-launch shader compilation from minutes to seconds and helping reduce in-game stutter, so players spend less time waiting, and more time playing.  Combined with growing support from game developers and publishers such as Electronic Arts, Embark, Tencent and Ubisoft joining KRAFTON, NetEase, Riot Games, Epic Games and more, players now have access to a deep and growing catalog of PC games on Windows on Arm, including Fortnite, Alan Wake 2 and the new World of Warcraft: Forever, and other top titles coming soon, like League of Legends, Valorant, PUBG: Battlegrounds, Genshin Impact and other top titles on Arm-based Windows PCs.  And we’re super excited to announce that in 2027, Call of Duty will be coming to RTX Spark™.   Windows is the home for PC gaming and together with our partners, we’re making Windows on Arm continue that heritage. That means more of the games players love, richer graphics, and less time waiting to get into the action.

Innovation built on a foundation of quality

Innovation only matters when the foundation underneath is great. Over the past year, the team has been hard at work improving Windows quality, and we continue to invest in everyday performance, reliability, and craftmanship that have meaningful impact on the way people experience Windows. This work has spanned giving customers more personalization and control in places like Windows updates, Start and taskbar, Search, the Windows Insider program, as well as deep investments with our partners to enhance driver quality, reliability and security with our ecosystem. It has also included significant efforts to make Windows the best destination for developers, including optimizations specific to the developer experience. In July we shared how this work continues as we deliver ongoing improvements customers can feel across each of these areas, which will only grow in importance with the new capabilities we’ve shared today. The way forward for Windows   This is the future we’re building: agents that can work securely on your behalf, intelligence that moves seamlessly between local and cloud, and devices that become more capable than ever before. At the same time, Windows will continue to get better at the everyday experiences people rely on — from Search to gaming and everything in between. From the edge to the frontier, Windows is the home for hybrid intelligence — helping individuals and organizations get more done, more efficiently, with security built in.  Explore the new Windows PCs powered by NVIDIA RTX Spark and join us at Microsoft Ignite in November to learn more about putting hybrid intelligence to work for your organization.

---

[1] Governance at scale may require additional services. [2] Windows Update required. Features will roll out over time, and availability and timing may vary by device, market, and silicon platform. [3] Windows Update required. Features will roll out over time, and availability and timing may vary by device, market, and silicon platform. [4] Availability and timing may vary by device, market, and silicon platform. [5] Compared with an Apple MacBook Pro 16-inch with M5 Pro and 64 GB of memory, tested Windows PCs with NVIDIA RTX Spark. [6] Based on Microsoft-commissioned testing conducted in September 2026, comparing a selection of preproduction Windows PCs with NVIDIA RTX Spark (64 GB) and a preproduction Microsoft Surface Laptop Ultra with NVIDIA RTX Spark (64 GB) against an Apple MacBook Pro 16-inch with M5 Pro (64 GB). Tested in llama.cpp using the Qwen3.5 27B model (Q4K-Medium) with a fixed 8,192-token prompt. Performance varies significantly by device configuration, usage and other factors. [7] Based on NVIDIA testing conducted in September 2026, comparing a selection of preproduction Windows PCs with NVIDIA RTX Spark (64 GB) and an Apple MacBook Pro 16-inch with M5 Pro (64 GB). Tested in ComfyUI using FLUX.2 Klein 4B at NVFP4 precision, four sampling steps, and an output resolution of 1024 × 1024. Performance varies by device. [8] Based on NVIDIA testing conducted in September 2026, comparing a selection of preproduction Windows PCs with NVIDIA RTX Spark (64 GB) and an Apple MacBook Pro 16-inch with M5 Pro (64 GB). Tested in ComfyUI using LTX 2.3 22B at NVFP4 precision, 121 frames at 25 fps, eight steps, and an output resolution of 1280 × 720. Performance varies by device. [9] Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip with up to 748 GB of coherent memory and 20 petaflops of FP4 AI compute, it delivers data-center-level performance, supporting large AI models up to 1 trillion parameters.

Hello Windows Insiders, Search is one of the most familiar ways to get around a Windows PC. It is how you may open an app, find a file, change a setting, or simply look for information. As Insiders, you have also been clear about what you expect from it: search should be fast, focused, and help you find what you need without friction. In July, we shared how we are improving the Windows Search Box with less clutter and more control. We introduced a calmer home screen, removed promotional content from web results, and gave you more choice over whether web and Microsoft Store suggestions appear. Today, we’re carrying those investments forward with a new Windows Search experience built on WinUI 3. This modern foundation makes Windows Search faster and more efficient with resources, while delivering an even more streamlined and focused design. [caption id="attachment_179190" align="aligncenter" width="2560"] Windows Search showing recent files, apps, settings, and an inline action to turn on do not disturb[/caption] Beginning 10/7, we’re rolling out this new experience on the taskbar for Windows Insiders. We are sharing this version early and intentionally so your feedback can help us fine-tune the experience, improve result relevance, identify stability issues across real-world hardware and workflows, and shape the UX before we expand the preview. Below, we’ll share what you can try today in the Experimental channel, what’s coming later this year, and what we’re building toward in the year ahead. First, this new Windows Search experience is built for speed. It’s built on WinUI 3 and is part of our broader investment in modernizing our core Windows experiences. Early testing shows substantial improvements in performance and memory usage, so it feels faster and more efficient from the moment you open it. Next, we’ve streamlined the UI into a single list to make it easier to scan and focus on what matters. Data shows that nearly all clicks in the current UX come from the results list on the left, rather than the large preview pane on the right. Informed by this behavior and your feedback, the new design brings results together in a single, easy-to-scan view that reduces visual clutter and puts the information you need in one place. We’re also improving how Windows Search understands what you’re looking for, including better matching for typos and synonyms. We’ll continue fine-tuning these improvements over the coming weeks. Later this year, inline previews and richer file details will make it easier to differentiate similar results and give you more context before opening a file. [caption id="attachment_179199" align="aligncenter" width="2560"] Windows Search showing inline previews for a PowerPoint presentation, a Word document, and a photo directly in search results[/caption] Rich inline web previews will bring answers to queries like “weather in San Diego” or “How tall is the Eiffel Tower?” directly into Windows Search, helping you get useful information without leaving the experience. [caption id="attachment_179201" align="aligncenter" width="2560"] Windows Search showing rich web previews for weather in San Diego, a stock price, and the height of the Eiffel Tower[/caption] You’ll also be able to narrow results with filters when exploring a category like files. These are experiences you may already be familiar with today but will be brought together in a more streamlined, single-pane layout. Lastly, we are expanding what search can help you accomplish. Search has always been about quickly helping you find what you need, and we’re turning Windows Search into a more capable starting point for getting things done in Windows with inline actions.

Actions: From finding to doing

We’re introducing inline actions to help you finish tasks right where you started, without opening another app or navigating through additional screens. We’re starting with familiar tasks where saving a few steps can make a meaningful difference in everyday use. Here are a few things to try:

• Turn something on or off. Try phrases such as “switch to dark mode," “turn on night light,” or “turn on Bluetooth.”
• Make a quick adjustment. Try “dim my screen” or “mute.”
• Manage your workspace. Try “minimize all my windows,” or “arrange windows on screen.”

[caption id="attachment_179195" align="aligncenter" width="1920"] Windows Search demonstrating an inline action to change the system theme to dark mode.[/caption] [caption id="attachment_179228" align="aligncenter" width="1920"] Windows Search demonstrating an inline action to snap Edge and Word[/caption] Coming soon, you’ll also see phone actions to complete quick tasks with your Phone Link connected device. [caption id="attachment_179197" align="aligncenter" width="1920"] Windows Search demonstrating an inline action to send a text on a Phone Link connected device[/caption] We’re starting with the actions people rely on most and will continue building on that foundation over time. Every piece of feedback helps us improve how search understands intent and expand the range of tasks you can complete directly from Windows Search.

What's ahead

Looking to next year, work is already underway to bring the new Windows Search experience to the Start menu and have it adapt naturally to different taskbar positions and Windows configurations. In our initial previews, you’ll continue to see the existing search experience when searching from the Start menu while we learn and refine the new experience with your feedback. Our goal is to make Windows Search more capable, with more natural ways to express what you need, richer information in context, and more ways to get things done directly from search – including thoughtful integrations with experiences like Copilot. Together, these improvements will help reduce the distance between finding something and getting it done, while preserving the speed and familiarity you expect today.

Try the new Windows Search and give us feedback

This search experience with inline actions is gradually rolling out to Windows Insiders starting today in the Experimental channel, with English as the supported language for this initial release. If you’d like to preview it sooner, you can also enable this experience in Settings > Windows Update > Windows Insider Program > Feature flags. After upgrading to build 26340.9616 in the Experimental channel, or build 28120.3181 in the Experimental (26H1) channel, you’ll need to click the "Check for updates" button in Windows Update, and restart your device before the ‘New Windows Search Experience’ feature flag is visible. If you are moving a device to a Windows Insider build for the first time, please see Get started with the Windows Insider Program - Windows Insider Program | Microsoft Learn. Share your feedback in Feedback Hub (WIN + F) under Desktop Environment > Search. Your feedback will help us tune the experience and shape what comes next. Anshul