Phi

Microsoft's family of high-performance Small Language Models (SLMs): engineered for cost-efficiency and on-device deployment.

Phi is Microsoft's open family of Small Language Models (SLMs) designed to deliver capabilities rivaling much larger models, but with a significantly smaller footprint. Models like Phi-3-mini (3.8B parameters) and Phi-4-reasoning (14B parameters) consistently outperform models of the same size and the next size up on benchmarks for language, reasoning, and coding. This efficiency is achieved through training on curated, 'textbook-quality' data. The result is a highly capable, cost-effective solution perfect for low-latency, resource-constrained environments: think fast, dependable AI running directly on-device or at the edge.

https://azure.microsoft.com/en-us/products/ai-services/azure-ai-studio/model-catalog/