No NPU Models available

[darylgraves-stantec]

Foundry Version: 0.8.119
Intel NPU Driver: 32.0.100.4512

When running foundry model list I can only see GPU and CPU models. I've tried upgrading NPU driver just in case.

If I find a model on the website and paste in the command (e.g foundry model run qwen2.5-coder-7b-instruct-openvino-npu:2) I get this:

Exception: ❌ Model <qwen2.5-coder-7b-instruct-openvino-npu:2> was not found in the catalog or local cache.
🔍 Did you mean: `foundry model run qwen2.5-coder-7b-instruct-generic-gpu:4`?
👉 Try `foundry model list` to see available models.
👉 Or check for typos in the model name.

Have the NPU models been removed? Thanks!

[samuel100]

The models are there. Using version 0.8.119, I see:

Is your device going through an OpenVino update (do you see it in Windows Update)?

[darylgraves-stantec]

Thanks for the quick response! This is what I see. I am not seeing any "registered the following EPs"

I am not seeing anything in Windows Update although this is a work computer so they are usually managed in other ways.

I updated the NPU drivers and rebooted today.

[natke]

Can you please share your Windows version and the foundry local logs (~/.foundry/logs)

[darylgraves-stantec]

This is my try this morning. I'm guessing my company's network settings are blocking it?

2026-03-12 08:46:33.494 +00:00 [INF] Starting Foundry Local CLI with ''
2026-03-12 08:46:41.576 +00:00 [INF] Starting Foundry Local CLI with 'model list'
2026-03-12 08:46:41.584 +00:00 [INF] Fetching model list from Azure Foundry catalog...
2026-03-12 08:46:41.821 +00:00 [INF] Timeout connecting to service
System.TimeoutException: The operation has timed out.
   at System.IO.Pipes.NamedPipeClientStream.ConnectInternal(Int32, CancellationToken, Int32) + 0xd5
   at System.Threading.ExecutionContext.RunFromThreadPoolDispatchLoop(Thread, ExecutionContext, ContextCallback, Object) + 0x3d
--- End of stack trace from previous location ---
   at System.Threading.ExecutionContext.RunFromThreadPoolDispatchLoop(Thread, ExecutionContext, ContextCallback, Object) + 0x70
   at System.Threading.Tasks.Task.ExecuteWithThreadLocal(Task&, Thread) + 0x6a
--- End of stack trace from previous location ---
   at Microsoft.Neutron.Rpc.Client.RpcSessionExtensions.<CreatePipeRpcSessionAsync>d__2`1.MoveNext() + 0x6f
--- End of stack trace from previous location ---
   at Microsoft.AI.Foundry.Local.Common.ServiceManagement.<ConnectClientAsync>d__8.MoveNext() + 0x6f
--- End of stack trace from previous location ---
   at Microsoft.AI.Foundry.Local.Common.ServiceManagement.<CheckIsRunning>d__9.MoveNext() + 0x5f
2026-03-12 08:46:41.823 +00:00 [INF] Starting service <C:\Program Files\WindowsApps\Microsoft.FoundryLocal_0.8.119.102_x64__8wekyb3d8bbwe\Inference.Service.Agent.exe --urls="http://127.0.0.1:0/" --OpenAIServiceSettings:ModelDirPath="C:\Users\dagraves\.foundry\cache\models" --JsonRpcServer:Run=true --JsonRpcServer:PipeName="inference_agent" --Logging:LogLevel:Default="Information">
2026-03-12 08:46:42.334 +00:00 [INF] Found service endpoints: http://127.0.0.1:52802
2026-03-12 08:46:42.334 +00:00 [INF] Service is started on http://127.0.0.1:52802/, PID 6184!
2026-03-12 08:46:42.335 +00:00 [INF] Command:ModelInit Status:Success Direct:True  Time:750ms
2026-03-12 08:46:42.335 +00:00 [INF] Command:ServiceStart Status:Failure Direct:False  Time:751ms
2026-03-12 08:46:42.335 +00:00 [INF] Checking EP autoregistration status...
2026-03-12 08:46:42.378 +00:00 [INF] Processing EP autoregistration status...
2026-03-12 08:46:42.378 +00:00 [INF] Command:ServiceStatus Status:Success Direct:False  Time:42ms
2026-03-12 08:46:42.378 +00:00 [INF] UserAgent:foundry-local-CLI/0.8.119 Command:ServiceStatus Status:Success Direct:True Time:0ms
2026-03-12 08:46:42.379 +00:00 [INF] UserAgent:foundry-local-CLI/0.8.119 Command:AutoRegisterCertifiedEps Status:Success Direct:True Time:0ms
2026-03-12 08:46:42.380 +00:00 [INF] Reporting progress of running EP download
2026-03-12 08:46:42.388 +00:00 [INF] Provider 1/1: OpenVINOExecutionProvider is in NotPresent state. Downloading and ensuring provider is ready.
2026-03-12 08:46:53.101 +00:00 [INF] Provider OpenVINOExecutionProvider download and ensuring ready attempt: Failure
2026-03-12 08:46:53.113 +00:00 [INF] Download attempt for OpenVINOExecutionProvider unsuccessful. Skipping all EP downloads. 
2026-03-12 08:46:53.118 +00:00 [INF] Finished attempt to autoregister certified EPs; finished in 10806ms
2026-03-12 08:46:53.119 +00:00 [INF] Failed to download or register the following EPs: OpenVINOExecutionProvider. Will try installing again later.
Valid EPs: CPUExecutionProvider, WebGpuExecutionProvider
2026-03-12 08:46:53.180 +00:00 [INF] Command:ServiceAutoRegister Status:Failure Direct:False  Time:10801ms
2026-03-12 08:46:53.181 +00:00 [INF] Creating new task to ensure and autoregister certified execution providers
2026-03-12 08:46:53.181 +00:00 [INF] Created task to ensure and autoregister certified execution providers
2026-03-12 08:46:53.181 +00:00 [INF] Attempt 2: Autoregistration of certified execution providers in progress.
2026-03-12 08:46:53.182 +00:00 [INF] Started autoregistering certified EPs
2026-03-12 08:46:53.182 +00:00 [INF] Device: CPU,EPs: CPUExecutionProvider
2026-03-12 08:46:53.182 +00:00 [INF] Device: GPU,EPs: WebGpuExecutionProvider
2026-03-12 08:46:53.216 +00:00 [INF] Provider 1/1: OpenVINOExecutionProvider is in NotPresent state. Downloading and ensuring provider is ready.
2026-03-12 08:46:54.023 +00:00 [INF] Model Phi-4-reasoning-generic-cpu:1 does not have a valid prompt template.
2026-03-12 08:46:54.215 +00:00 [INF] Model Phi-4-reasoning-generic-gpu:1 does not have a valid prompt template.
2026-03-12 08:46:54.215 +00:00 [INF] Total models fetched across all pages: 41
2026-03-12 08:46:54.215 +00:00 [INF] UserAgent:foundry-local-CLI/0.8.119 Command:ModelList Status:Success Direct:True Time:1034ms
2026-03-12 08:46:54.223 +00:00 [INF] Command:ModelList Status:Success Direct:True  Time:12639ms
2026-03-12 08:46:54.224 +00:00 [INF] Stream disconnected

[natke]

Can you share your Windows version please?

[darylgraves-stantec]

Sorry! Yes I forgot. It's Window 11 Enterprise 24H2, OS Build 26100.7840.

[natke]

Thank you. Can you please create a Windows feedback issue, following these instructions? https://learn.microsoft.com/en-us/windows/ai/new-windows-ml/execution-provider-errors#how-to-file-feedback

[natke]

If you share the feedback link, I will follow up internally

[darylgraves-stantec]

Here we are: https://aka.ms/AA106wg4

Sorry, there may be two reports in your system which are the same. I didn't realise I had to be logged in to make a shareable link.

I think somewhere in my troubleshooting I read that it uses Windows Updates in some form? I do wonder if this is linked to the fact that my company manage Windows Updates via SCCM so we are unable to reach out to the Windows Update URLs directly. That said, I'm not sure why NPU models would be hidden behind Windows Update URLs when others are available without.

Thanks for all your help!

[natke]

Thanks for that @darylgraves-stantec, yes the NPU availablility is gated by Windows update. The accelerator that allows the NPU to run gets downloaded, installed and registered with Foundry Local by windows update. If that fails, then the NPU models are not available to Foundry Local. If your company policy is preventing this, the Windows team should be able to tell this from the feedback hub report

[natke]

Unfortunately it does look like your enterprise policy is blocking the download of the NPU execution provider. We are working on solutions to solving this issue. But you are blocked from running NPU models for now

[darylgraves-stantec]

Thanks Nat, do you think that will change in the near future? Or is it worth me just closing the ticket for now and finding an alternative?

[MarioVolarevic]

I'm having the exact same issue with VS code AI toolkit.
During setup, it triggers the Windows Update and fails to load the provider because the company is blocking these updates.
So it would be really nice if there could be some alternative way to configure download from somewhere else.