What KB5136900 actually delivers
Here are the facts from Microsoft's support page:
- Component: NVIDIA TensorRT-RTX Execution Provider
- New version: 2.2611.1.0
- Published: November 19, 2026
- Replaces: KB5132812
- Delivery: Downloads and installs automatically through Windows Update
- Prerequisite: The latest cumulative update for Windows 11 version 24H2, 25H2 or 26H2
- Update history entry: "Windows ML Runtime NVIDIA TensorRT-RTX Execution Provider Update (KB5136900)"
The page doesn't say much about the changes themselves. It says only that the release "includes improvements to the execution provider component." Microsoft lists no changelog, bug fixes, security fixes or performance figures. Any claim that this update speeds up your AI workloads by some amount is not based on anything Microsoft has published.
Who gets it
The edition list is split:
| Windows 11 version | Editions listed | Availability |
|---|---|---|
| 24H2 | Enterprise and Education, Enterprise multi-session, IoT Enterprise | Not offered until the next servicing period, on or after December 8, 2026 |
| 25H2 | All editions | Through Windows Update |
| 26H2 | All editions | Through Windows Update |
If you run 24H2 and don't see the update yet, Windows Update isn't broken. Microsoft is deliberately holding the update back for that version until December.
Section summary: KB5136900 is a component update with no detailed changelog. 25H2 and 26H2 devices can get it through Windows Update now, and 24H2 devices get it after December 8.
What an execution provider does
Microsoft describes the TensorRT-RTX provider as an ONNX Runtime and Windows ML execution provider that speeds up ONNX model inference on NVIDIA RTX GPUs in end-user PC scenarios. It uses NVIDIA's TensorRT for RTX runtime to build and run inference engines tuned for the RTX hardware in the machine.
ONNX Runtime can't get the best performance out of every chip on its own, so it hands supported parts of a model to hardware-specific backends. An execution provider works like a specialist subcontractor. CPU work stays with the CPU, and the RTX provider takes the work an RTX GPU handles well.
The ONNX Runtime documentation describes how TensorRT RTX works. At inference time, the EP context model is loaded and TensorRT RTX dynamically compiles the binary blob (engine) to optimize it for the exact GPU hardware being used. According to that documentation, this approach is designed to build RTX hardware-optimized engines on end-user devices in seconds. This is why the provider is aimed at consumer PCs, where developers can't build a separate engine for every GPU their users might own.
On the developer side, the same ONNX Runtime documentation says the built-in TensorRT RTX Execution Provider in the ONNX Runtime repository is deprecated, and the project strongly recommends using the standalone EP ABI plugin instead. NVIDIA's GitHub repository describes that plugin as a component built as a standalone plugin that implements the ORT Plugin EP ABI and leverages TensorRT for RTX to accelerate ONNX models on RTX GPUs (Ampere and later). The provider can be serviced through KB updates partly because it ships as a separate plugin.
Section summary: The provider is the layer that compiles and runs AI models on RTX GPUs for apps built on Windows ML or ONNX Runtime. It is not a GPU driver and not a Windows feature update.
Hardware and driver requirements
KB5136900 doesn't list any new hardware requirements. Microsoft's Windows ML documentation on Learn lists the following requirements for NvTensorRtRtxExecutionProvider:
- An NVIDIA GeForce RTX 30-series or newer GPU
- NVIDIA driver version 32.0.15.5585 or later (minimum recommended)
- CUDA 12.5
These requirements come from the Windows ML documentation and are not new with this KB. The same documentation says these downloadable providers are available on Windows 11 version 24H2 (build 26100) or later, depending on device and driver compatibility. NVIDIA's own repository also says the provider supports RTX GPUs based on Ampere and later architectures (NVIDIA GeForce RTX 30xx and above).
If your PC has a GTX card or an RTX 20-series card, the provider isn't built for that hardware.
One of five vendor updates released the same day
Microsoft didn't release KB5136900 on its own. Microsoft's release history of AI updates for Windows 11 versions 24H2, 25H2 and 26H2 shows five execution provider updates dated November 19, 2026: Version2.2611.1.0KB5136903 | Version2.2611.0.0KB5136902 | Version2.2611.1.0KB5136901 | Version2.2611.1.0KB5136900 | Version2.2611.1.0KB5136899. In order, those columns are AMD MIGraphX, AMD Vitis, Intel OpenVINO, NVIDIA TensorRT-RTX and Qualcomm QNN.
The table also shows a second NVIDIA package for a different Windows version. Windows 11 version 26H1 has its own servicing track, and its NVIDIA entry is Version2.2611.1.0KB5136905. The two NVIDIA updates share a version number but have different KB numbers:
| Windows 11 version | NVIDIA TensorRT-RTX KB | Provider version |
|---|---|---|
| 24H2, 25H2, 26H2 | KB5136900 | 2.2611.1.0 |
| 26H1 | KB5136905 | 2.2611.1.0 |
Admins should check that their compliance reports match each machine's OS version to the right KB. Looking for KB5136900 on a 26H1 device will make a fully updated PC look out of date.
Section summary: This is a coordinated monthly release across five vendors. NVIDIA's November build uses two KB numbers that depend on the Windows 11 version.
How to check that KB5136900 is installed
- Open Settings.
- Go to Windows Update > Update history.
- Look for Windows ML Runtime NVIDIA TensorRT-RTX Execution Provider Update (KB5136900).
If it isn't listed, check these first:
- Are you on 24H2? The update isn't offered on 24H2 until on or after December 8, 2026.
- Do you have the latest cumulative update? Microsoft lists it as a prerequisite.
- Is the PC managed? Microsoft's Windows ML troubleshooting guidance says enterprise policies can block execution providers from downloading on managed devices. Ask your IT team whether a policy restricts component downloads or access to Windows Update.
- Are updates paused, or is a restart pending? The same guidance lists both as known causes of provider download failures, at least for older Windows ML versions.
A note for developers: two delivery paths
Developers should keep two delivery paths separate. Microsoft's KB says KB5136900 arrives through Windows Update. However, the Windows ML troubleshooting documentation says the way an app acquires execution providers changed in Windows ML 2.5.96. From that version onward, providers come through the Microsoft Store. Versions 2.4.89 and earlier used Windows Update. Microsoft recommends that developers use Windows ML 2.5.96 or later for the most reliable provider acquisition.
So if a Windows ML app can't load the NVIDIA provider, KB5136900 isn't necessarily the cause. Check which Windows ML version the app ships with, then troubleshoot the delivery path that version uses.
Analysis: should you care?
For most RTX owners, nothing needs to be done. The update installs automatically and doesn't change anything you'll notice. Its main value is that Microsoft and NVIDIA can now update the AI inference layer every month without waiting for a feature update or a driver release.
The downside is that Microsoft says very little about what changes. "Improvements" doesn't help an admin judge risk or a developer chase a regression. NVIDIA's TensorRT for RTX download page lists upstream work such as Qwen3.5 dense models are supported through Windows ML with the TensorRT-RTX execution provider. That page doesn't tie the work to version 2.2611.1.0, so don't assume any of it ships in KB5136900.
Key points
- Home users with RTX 30-series or newer GPUs: Let Windows Update install it. No action is needed.
- 24H2 Enterprise, Education and IoT fleets: Expect the update on or after December 8, 2026, and keep cumulative updates current.
- 25H2 and 26H2: The update is available now through Windows Update.
- Compliance reporting: KB5136900 covers 24H2, 25H2 and 26H2. KB5136905 is the matching update for 26H1.
- Developers: Use Windows ML 2.5.96 or later, and check provider availability in code instead of assuming a particular KB is installed.
References
- KB5136900: Nvidia TensorRT-RTX Execution Provider update (version 2.2611.1.0) | Microsoft Support Microsoft Support · 2026-11-19T00:00:00+00:00
- KB5136900: Nvidia TensorRT-RTX Execution Provider update (version 2.2611.1.0) | Microsoft Support support.microsoft.com
- GitHub - NVIDIA/TensorRT-RTX-EP-ABI: The NVIDIA TensorRT RTX Execution Provider (EP) is an inference deployment solution designed specifically for NVIDIA RTX GPUs, optimized for client-centric use cases. · GitHub github.com