Editja

Choose your Windows AI solution

The Windows AI landscape has grown quickly and the terminology can be hard to navigate. This page cuts through it: find your scenario, pick your starting point.

Options

  • Ready-to-use AI models and APIs — Models distributed and managed by Microsoft; minimal code, no ML expertise needed.
    • Windows AI APIs — LLM (Phi Silica), imaging models, OCR, semantic search, and more. Most APIs require a Copilot+ PC; some (Phi Silica on GPU, Speech Recognition, Video Super Resolution) also run on other Windows 11 hardware—see the supported hardware table.
    • Foundry Local — 20+ open-source LLMs and speech models via an OpenAI-compatible API. Any Windows hardware.
  • Run any model locally — Bring your own ONNX model with full control over the inference pipeline.
    • Windows ML — Hardware-accelerated inference on CPU, GPU, or NPU. Any Windows hardware.
  • Other
    • Microsoft Foundry — Cloud-hosted frontier models (GPT-4o, DALL-E, etc.) via REST API. Combine with Foundry Local for on-device/cloud fallback.

Terminology decoder

The Windows AI space has gone through rapid rebranding. Here's a translation table for terms you might encounter:

Term you've seen What it means now
Copilot Runtime APIs Old name (2024) for Windows AI APIs. Same functionality, renamed.
Windows Copilot Runtime Old umbrella term (2024) for the AI features now called Microsoft Foundry on Windows.
Windows AI Foundry Old umbrella term (2025) for the AI features now called Microsoft Foundry on Windows.
Microsoft Foundry on Windows Current umbrella brand covering Windows AI APIs + Foundry Local + Windows ML.
Microsoft Foundry Microsoft's cloud-based AI platform. Different product, different team, similar name.
Phi Silica The specific Phi model optimized for and built into Windows on Copilot+ PCs (NPU). Also available on non-Copilot+ PCs with a supported NVIDIA or AMD GPU. Accessed via Windows AI APIs.
Phi (general) Microsoft's family of small language models. Phi-4-mini etc. are available on Azure AI and via Foundry Local. Phi Silica is the Windows inbox version, optimized for NPU and supported GPUs.
Windows ML (old) Legacy WinRT-based inference API, inbox since Windows 10 1809. Still works; no new investment.
Windows ML (new) New ONNX Runtime-based NuGet package. Current and actively developed.
DirectML No longer being actively developed (in sustained engineering). Low-level DirectX 12 ML API for GPU/NPU acceleration.
Windows ML IHV-specific Execution Providers The replacement for DirectML, achieving higher performance by working natively with Windows hardware.
Copilot+ PC A PC category defined by hardware: NPU with 40+ TOPS, 16GB+ RAM, specific SoCs. Required for most Windows AI APIs; a growing number of APIs (Phi Silica on GPU, Speech Recognition, Video Super Resolution) also run on non-Copilot+ Windows 11 PCs. Not required for Foundry Local or Windows ML.
NPU Neural Processing Unit — dedicated AI acceleration hardware in Copilot+ PCs. On a Copilot+ PC, Windows AI APIs route inference through the NPU automatically; some APIs also support running on GPU or CPU on non-Copilot+ devices.

Combine options in the same app

These options aren't mutually exclusive. A typical pattern for a resilient AI feature falls back across three tiers: first try Windows AI APIs for the fastest, most optimized experience on supported hardware; if unavailable, fall back to Foundry Local to run an open-source model on any Windows hardware; and if that's also unavailable, fall back to Azure AI in the cloud. The following example demonstrates this pattern:

// 1. Try Windows AI APIs (Copilot+ PC, or supported GPU/CPU for some APIs)
using Microsoft.Windows.AI;
using Microsoft.Windows.AI.Text;

var readyState = LanguageModel.GetReadyState();
if (readyState == AIFeatureReadyState.NotReady)
{
    var deploymentResult = await LanguageModel.EnsureReadyAsync();
    if (deploymentResult.Status == AIFeatureReadyResultState.Success)
    {
        readyState = LanguageModel.GetReadyState();
    }
    else
    {
        // Optional: inspect deploymentResult.ExtendedError for diagnostics.
        // Treat as unavailable so we fall through to Foundry/Azure.
        readyState = AIFeatureReadyState.NotSupportedOnCurrentSystem;
    }
}

if (readyState == AIFeatureReadyState.Ready)
{
    // Use Phi Silica via Windows AI APIs
    using LanguageModel languageModel = await LanguageModel.CreateAsync();
}
// 2. Fall back to Foundry Local (any hardware)
else if (await foundryClient.IsModelAvailableAsync("phi-4-mini"))
{
    // Use Foundry Local OpenAI-compatible API
}
// 3. Fall back to Azure AI (always available)
else
{
    // Use Azure OpenAI
}

This pattern gives Copilot+ users the best experience while keeping the feature working on all hardware.

Still unsure?

  • Last updated on