Phi-4-mini-onnx-qnn-npu

This is a quantized INT4 model based on Qualcomm Copilot+PC NPU. You can deploy it on Copilot+PC devices.

Note: This is unoffical version,just for test and dev.

Installation

Please use Windows 11 + Copilot, install Foundry Local 0.6.87

winget install Microsoft.FoundryLocal

Download Models

Please use Huggingface CLI download this model

hf download lokinfey/Phi-4-mini-onnx-qnn-npu --local-dir Your Phi-4-mini-onnx-qnn-npu Path

add inference_model.json to Your Phi-4-mini-onnx-qnn-npu Path/model

Deployment

foundry service stop

foundry cache cd Your Phi-4-mini-onnx-qnn-npu

You can run foundry cache ls to check model status

Running

foundry model run Phi-4-mini-qnn-npu