Using Phi-3 Medium 4K on Microsoft Foundry
Implementation guide · Phi-3 · Microsoft Research
Quick Start
- 1
- 2Use the Microsoft Foundry SDK or REST API to call
phi-3-medium-4k— see the documentation for request format. - 3
Code Examples
About Microsoft Foundry
Microsoft Foundry offers a comprehensive platform-as-a-service for enterprise AI operations. It provides multiple deployment options including Serverless APIs (pay-as-you-go), Global Standard (shared managed capacity), Provisioned Throughput Units (reserved capacity), batch processing, and bring-your-own model deployments. The platform features a unified control plane for models, agents, tools, and observability. Its Agent Service enables building and deploying AI agents with built-in tracing, monitoring, and governance. Evaluation and monitoring tools assess model performance, safety, and groundedness. Foundry supports seamless upgrades from Azure OpenAI with non-destructive migration, maintaining existing deployments while unlocking multi-provider model access and advanced platform capabilities.
Microsoft Foundry is a unified Azure platform-as-a-service offering for enterprise AI operations, model builders, and application development. It provides access to over 1,900 models from Microsoft, OpenAI, Anthropic, Mistral, xAI, Meta, DeepSeek, Hugging Face, and more. Foundry unifies agents, models, and tools under a single management grouping with built-in enterprise-readiness capabilities including tracing, monitoring, evaluations, and customizable enterprise setup configurations.
Pricing on Microsoft Foundry
| Type | Price (per 1M) |
|---|---|
| Input tokens | $0.45 |
| Output tokens | $1.35 |
Capabilities
About Phi-3 Medium 4K
The Phi-3 Medium 4K, developed by Microsoft, is a state-of-the-art large language model with 14 billion parameters. It is engineered for efficiency across various tasks, particularly excelling in reasoning capabilities. This model is designed to handle 4,096 token context lengths, allowing for the processing of longer input sequences. Leveraging a dense, decoder-only Transformer architecture, it incorporates techniques like supervised fine-tuning and direct preference optimization to align with human preferences and safety standards. The model supports multilingual data, although it is primarily trained in English. Its lightweight nature allows for deployment on diverse hardware platforms, making it accessible and versatile for both commercial and research purposes. Safety measures are embedded, although further precautions are advised for applications with higher risks.