Microsoft Foundry Local: Unified On-Device AI Runtime Replaces Ollama, llama.cpp, whisper.cpp

GitHub·medium signal

Microsoft's Foundry Local (2,059 stars) provides a unified local AI runtime with one SDK handling both chat and audio inference, with automatic hardware acceleration (NPU > GPU > CPU). All inference runs on-device with zero network latency, no per-token costs, and full offline capability. A March 23 hands-on guide pairs it with the Microsoft Agent Framework for building real-world AI automation, positioning it as Microsoft's answer to Ollama for the agent-native developer workflow.