// about Microsoft Phi
Microsoft's Phi series (Phi-2, Phi-3, Phi-4) are small language models that consistently outperform much larger models on reasoning and coding benchmarks, validating Microsoft Research's thesis that data quality matters more than model size. The Phi-3-mini (3.8B) runs on smartphones, while Phi-3-medium and Phi-4 deliver near-GPT-4-class performance at a fraction of the cost. Widely available via Azure, Ollama, and Hugging Face, Phi models are a go-to choice for edge deployment and cost-sensitive applications.