KB5075032 landed on January 13, 2026: Microsoft has published a new Phi Silica AI component update for Intel-powered Copilot+ PCs — version 1.2511.1326.0 — delivered through Windows Update for Windows 11 (24H2 and 25H2). Introduction — why this update matters
Phi Silica is Microsoft’s on-device, Transformer-based “small language model” (SLM) that’s been tuned specifically to run efficiently on the Neural Processing Units (NPUs) inside Copilot+ PCs. That combination — a compact but capable model plus NPU offload — is designed to give low-latency, privacy-friendly language features locally on the device (for things like Click-to-Do, summarization, and other Copilot experiences), and Microsoft treats Phi Silica as a platform-level AI component that is updated separately from regular Windows cumulative updates. This particular release (KB5075032) is the Intel-targeted deployment of the newer Phi Silica build (1.2511.1326.0). It replaces the previous Intel build that Microsoft shipped under KB5072642 (version 1.2511.1196.0). If your device is an Intel-based Copilot+ PC running Windows 11 24H2 or 25H2 and you rely on built-in Copilot features, this update is the Intel-side rollout of that platform component and will appear automatically through Windows Update. At-a-glance: the official facts
- Update name and KB: KB5075032 — Phi Silica AI component update (version 1.2511.1326.0) for Intel-powered systems.
- Applies to: Windows 11, version 24H2 and Windows 11, version 25H2 (Copilot+ PCs only).
- Delivery: automatic via Windows Update (no separate download required for most consumer devices).
- Prerequisite: the latest cumulative update for Windows 11 24H2 or 25H2 must be installed.
- Replacement: this update replaces the earlier Intel release (KB5072642, v1.2511.1196.0).
Phi Silica is part of Microsoft’s family of compact, NPU-targeted models that are intended to provide many of the conversational and reasoning capabilities you’d expect from a larger LLM, but optimized for power and latency so they can run locally on Copilot+ hardware. The Windows AI APIs and the Windows App SDK expose Phi Silica functionality to apps (subject to Microsoft’s limited-access policies for some features), and the model supports techniques such as speculative decoding to accelerate generation by combining a fast draft model with a verifier. Microsoft’s public developer and product posts describe Phi Silica as an NPU-tuned model that supports multimodal extensions (image understanding) and reasoning variants (Phi‑4 family) that are being made available across device classes over time. Why the version bump matters (1.1196.0 → 1.1326.0)
Microsoft’s KBs show a clear sequence of component updates: November 2025 Intel/AMD releases referenced version 1.2511.1196.0 (KB5072642 for Intel and KB5072643 for AMD). KB5075032 publishes a later Intel-targeted build (1.2511.1326.0) in January 2026. The change in the third numeric block (1196 → 1326) indicates an incremental platform-level release — typically bug fixes, performance or compatibility tweaks, or enabling new NPU-side improvements — rather than a complete rearchitecture. The KB itself is intentionally concise about “what changed” (it summarizes scope, prerequisites, and replacement information) rather than providing a line-by-line changelog; Microsoft often publishes more detailed engineering notes for developers on Learn or Windows blogs when component behavior or APIs change. What Microsoft explicitly says you should expect
- The update will be installed automatically via Windows Update for eligible Copilot+ devices that meet the prerequisites. Check Settings > Windows Update > Update history to confirm the update shows up as “Phi Silica version 1.2511.1326.0 for Intel-powered systems (KB5075032)”.
- If you have Intel hardware and were on the earlier KB5072642 release, KB5075032 is the replacement for that earlier Intel package.
- Microsoft Learn (Windows AI / Phi Silica pages) describes Phi Silica as an NPU-tuned local language model, notes limited-access controls for some features, and documents the Windows AI APIs developers can use to call Phi Silica from apps. That page is the canonical place for API and developer guidance.
- The Windows Experience Blog and Azure blog posts provide product context (multimodal capabilities, Phi-4 reasoning family, NPU offload benefits and rollout timelines) and confirm the ongoing work to extend Phi Silica capabilities (vision, reasoning variants) across Snapdragon, Intel, and AMD Copilot+ NPUs. Use those posts for feature-level context rather than KB-level installation detail.
- Settings > Windows Update > Update history — look for “Phi Silica version 1.2511.1326.0 for Intel-powered systems (KB5075032)” under the Windows Update history list after the update installs. This is the method Microsoft documents in the KB.
- If you manage machines centrally (WSUS, SCCM/ConfigMgr, or Windows Update for Business), check your device update reports and the updates catalog entries for the KB number. (Microsoft surfaces the KB ID so IT tools can filter by KB.
- If you’re a consumer: you do not need to take explicit action. Windows Update will download and install the component on eligible Copilot+ devices; reboot when prompted so the new component is fully active. Confirm in Settings > Windows Update > Update history.
- If you’re an IT pro or admin: verify that devices targeted for Phi Silica updates have the required cumulative Windows 11 24H2/25H2 updates in place; set your Windows Update management policies accordingly so this component can be delivered. If you use ringed deployment, pilot on a small set of Intel Copilot+ devices before broad roll‑out.
- If you’re a developer: consult the Windows AI / Phi Silica documentation and the limited-access feature guidance on Microsoft Learn before relying on any LAF-protected functionality; expect platform component updates to be delivered by Windows Update and to affect on-device model behavior.
- Local execution: a key design point for Phi Silica is on-device inference (NPU offload), which keeps model execution and many inference signals local to the device unless the product experience is explicitly hybrid with cloud services. This can reduce telemetry of raw user prompts off-device, but the service interfaces and possible cloud augmentation scenarios are governed by Microsoft’s product and privacy policies. For precise privacy and telemetry details, review Microsoft’s privacy notices and product documentation for Copilot and Windows AI features.
- Limited Access Features (LAF): Microsoft’s Phi Silica APIs are documented as part of the Windows App SDK and some features are gated behind limited access controls; developers and OEMs should follow the LAF process if a specific capability is required.
- If the update doesn’t appear automatically: make sure the device has the latest cumulative Windows 11 update installed (KB prerequisite stated in the KB). Force a manual check: Settings > Windows Update > Check for updates. If the device is enrolled in a managed update service (WSUS, MEM, WUfB), confirm enterprise policies don’t defer or block components.
- If an app’s behavior changes after the update: because Phi Silica is a model component used by system-level AI experiences and some apps, a component-level update can change generation behavior (performance, subtle quality changes). For reproducibility, collect logs, note OS and Phi Silica versions, and work with Microsoft support or the Microsoft Tech Community if you need deeper investigation. Consider rolling the update out in phases for production fleets.
Microsoft’s public posts explain that Phi Silica initially targeted NPUs in Snapdragon X-series Copilot+ devices and has been extended to other Copilot+ silicon over time; the company highlights multimodal work (vision + text) and new reasoning model variants (Phi‑4 family) that benefit from ONNX and NPU offload. Those posts explain the broader roadmap and the reason behind device-targeted component KBs (separate KBs for AMD, Intel, Qualcomm in different timelines). In short: expect hardware-specific KBs and staged rollouts while Microsoft harmonizes model flavors and NPU runtimes across OEM silicon. What’s not (yet) in the public KB
- The Microsoft KB for this component is intentionally focused and does not publish a detailed line-by-line changelog inside that page; it documents scope, the version number, replacement information, prerequisite instructions, and the update delivery mechanism. For deeper technical notes (API changes, expanded multimodal support, or model behavior changes), Microsoft publishes developer and product blog posts or Learn documentation; consult those channels if you need feature-level specifics.
- Pilot first: designate a small set of Intel Copilot+ devices for early installation and user acceptance testing. Monitor for any unexpected changes in Copilot experiences or app integrations that use Windows AI APIs.
- Confirm prerequisites: ensure each device has the latest Windows cumulative update for 24H2/25H2. Microsoft requires that the overall OS cumulative update is present before delivering the Phi Silica component.
- Document versions: when you approve and deploy, keep a record (KB number and Phi Silica version) to track compatibility with any app integrations or internal audits. The KB number (KB5075032) and version (1.2511.1326.0) are what you’ll see in Update history.
- 2025-11: Phi Silica version 1.2511.1196.0 appears in earlier KBs (KB5072642 for Intel and KB5072643 for AMD).
- 2026-01-13: Microsoft publishes KB5075032 — Phi Silica version 1.2511.1326.0 for Intel-powered systems (this KB replaces KB5072642 for Intel). If you are on an Intel Copilot+ PC, this is the new component build delivered via Windows Update.
- Microsoft Support KB: KB5075032 (this update’s official KB article, for confirmation and the exact wording used by Microsoft).
- Microsoft Learn: Windows AI / Get started with Phi Silica (developer APIs, limited-access notes, and technical background on how the model is exposed to apps).
- Windows Experience Blog and Azure Blog posts: product-level context about multimodal enablement, Phi variants (Phi‑4), and NPU offload rationale. These are useful when you want to understand the features Microsoft is enabling on Copilot+ PCs beyond the component update itself.
- Draft an example IT rollout checklist and a short pilot test plan tailored to Intel Copilot+ devices in your environment (WSUS, MEM, or unmanaged), or
- Check whether your specific device model is listed by the OEM as a Copilot+ device and confirm any OEM firmware/driver prerequisites that Microsoft or the OEM recommends before deploying Phi Silica updates.
Source: Microsoft Support KB5075032: Phi Silica AI component update (version 1.2511.1326.0) for Intel-powered systems - Microsoft Support
