Microsoft Phi Models

Small model series for efficient local inference and edge scenarios.

Latest: Phi‑3.5

Phi‑3.5 provides efficient inference for edge and local deployments with reasonable capabilities for assistants.

Strengths

  • Small models
  • Efficient inference
  • Edge deployment

Best For

  • On‑device assistants
  • Local tools
  • Resource‑constrained workloads

Quick Tips

  • Use quantized variants for mobile and edge devices.
  • Stream responses to improve perceived latency.
  • Pair with lightweight RAG for grounded outputs.