Support for Phi Silica

[nwillems94]

Is it possible to add support in foundry for Phi Silica, the AI model bundled with CoPilot+ PCs?

Suggested functionality would allow running the model,
foundry model run phi-silica
and adding it to the service
foundry model load phi-silica

A version of this functionality exists in the VS Code AI Toolkit, but it would be nice to have it bundled in foundry.

[leestott]

Hi @nwillems94 What is the recommended pattern right now

On Windows AI Copilot+ PCs

Use Windows AI APIs (Phi Silica via WCR)

Everywhere else

Use Foundry Local with:
Phi‑3.5‑mini
Phi‑4‑mini / Phi‑4
Qwen / DeepSeek / etc.

This gives you:

Maximum performance on NPUs where available
Portable fallback everywhere else
A clear separation of platform model vs portable model

References:

Windows AI APIs (Phi Silica): https://learn.microsoft.com/windows/ai/apis/phi-silica
Foundry Local models & CLI: https://learn.microsoft.com/azure/foundry-local/get-started