README.md

Phi Cookbook: Hands-On Examples wit Microsoft Phi Models

Phi na series of open source AI models wey Microsoft develop.

Phi na di most powerful and cost-effective small language model (SLM) now, get beta beta benchmarks for multi-language, reasoning, text/chat generation, coding, images, audio and oda scenarios.

You fit deploy Phi go cloud or edge devices, and you fit easy build generative AI applications wit limited computing power.

Make you follow these steps to start use these resource:

1. Fork di Repository: Click
2. Clone di Repository: git clone https://github.com/microsoft/PhiCookBook.git
3. Join Di Microsoft AI Discord Community and meet experts and fellow developers

🌐 Multi-Language Support

Supported via GitHub Action (Automated & Always Up-to-Date)

Arabic | Bengali | Bulgarian | Burmese (Myanmar) | Chinese (Simplified) | Chinese (Traditional, Hong Kong) | Chinese (Traditional, Macau) | Chinese (Traditional, Taiwan) | Croatian | Czech | Danish | Dutch | Estonian | Finnish | French | German | Greek | Hebrew | Hindi | Hungarian | Indonesian | Italian | Japanese | Kannada | Khmer | Korean | Lithuanian | Malay | Malayalam | Marathi | Nepali | Nigerian Pidgin | Norwegian | Persian (Farsi) | Polish | Portuguese (Brazil) | Portuguese (Portugal) | Punjabi (Gurmukhi) | Romanian | Russian | Serbian (Cyrillic) | Slovak | Slovenian | Spanish | Swahili | Swedish | Tagalog (Filipino) | Tamil | Telugu | Thai | Turkish | Ukrainian | Urdu | Vietnamese

You prefer to Clone for your machine?

Dis repository get 50+ language translations wey heavy for download size. To clone witout translations, use sparse checkout:

Bash / macOS / Linux:

git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git
cd PhiCookBook
git sparse-checkout set --no-cone '/*' '!translations' '!translated_images'

CMD (Windows):

git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git
cd PhiCookBook
git sparse-checkout set --no-cone "/*" "!translations" "!translated_images"

Dis one go give you everything wey you need to finish di course with fast download.

Table of Contents

• Introduction

  • Welcome to the Phi Family
  • Setting up your environment
  • Understanding Key Technologies
  • AI Safety for Phi Models
  • Phi Hardware Support
  • Phi Models & Availability across platforms
  • Using Guidance-ai and Phi
  • GitHub Marketplace Models
  • Azure AI Model Catalog

• Inference Phi in different environment

  • Hugging face
  • GitHub Models
  • Microsoft Foundry Model Catalog
  • Ollama
  • AI Toolkit VSCode (AITK)
  • NVIDIA NIM
  • Foundry Local

• Inference Phi Family

  • Inference Phi in iOS
  • Inference Phi in Android
  • Inference Phi in Jetson
  • Inference Phi in AI PC
  • Inference Phi with Apple MLX Framework
  • Inference Phi in Local Server
  • Inference Phi in Remote Server using AI Toolkit
  • Inference Phi with Rust
  • Inference Phi--Vision in Local
  • Inference Phi with Kaito AKS, Azure Containers(official support)

• Quantifying Phi Family

  • Quantizing Phi-3.5 / 4 using llama.cpp
  • Quantizing Phi-3.5 / 4 using Generative AI extensions for onnxruntime
  • Quantizing Phi-3.5 / 4 using Intel OpenVINO
  • Quantizing Phi-3.5 / 4 using Apple MLX Framework

• Evaluation Phi

  • Response AI
  • Microsoft Foundry for Evaluation
  • Using Promptflow for Evaluation

• RAG with Azure AI Search

  • How to use Phi-4-mini and Phi-4-multimodal(RAG) with Azure AI Search

• Phi application development samples

  • Text & Chat Applications

    • Phi-4 Samples
      • [📓] Chat With Phi-4-mini ONNX Model
      • Chat with Phi-4 local ONNX Model .NET
      • Chat .NET Console App with Phi-4 ONNX using Sementic Kernel
    • Phi-3 / 3.5 Samples
      • Local Chatbot in the browser using Phi3, ONNX Runtime Web and WebGPU
      • OpenVino Chat
      • Multi Model - Interactive Phi-3-mini and OpenAI Whisper
      • MLFlow - Building a wrapper and using Phi-3 with MLFlow
      • Model Optimization - How to optimize Phi-3-min model for ONNX Runtime Web with Olive
      • WinUI3 App with Phi-3 mini-4k-instruct-onnx -WinUI3 Multi Model AI Powered Notes App Sample
      • Fine-tune and Integrate custom Phi-3 models with Prompt flow
      • Fine-tune and Integrate custom Phi-3 models with Prompt flow in Microsoft Foundry
      • Evaluate the Fine-tuned Phi-3 / Phi-3.5 Model in Microsoft Foundry Focusing on Microsoft's Responsible AI Principles
      • [📓] Phi-3.5-mini-instruct language prediction sample (Chinese/English)
      • Phi-3.5-Instruct WebGPU RAG Chatbot
      • Using Windows GPU to create Prompt flow solution with Phi-3.5-Instruct ONNX
      • Using Microsoft Phi-3.5 tflite to create Android app
      • Q&A .NET Example using local ONNX Phi-3 model using the Microsoft.ML.OnnxRuntime
      • Console chat .NET app with Semantic Kernel and Phi-3

  • Azure AI Inference SDK Code Based Samples

    • Phi-4 Samples
      • [📓] Generate project code using Phi-4-multimodal
    • Phi-3 / 3.5 Samples
      • Build your own Visual Studio Code GitHub Copilot Chat with Microsoft Phi-3 Family
      • Create your own Visual Studio Code Chat Copilot Agent with Phi-3.5 by GitHub Models

  • Advanced Reasoning Samples

    • Phi-4 Samples
      • [📓] Phi-4-mini-reasoning or Phi-4-reasoning Samples
      • [📓] Fine-tuning Phi-4-mini-reasoning with Microsoft Olive
      • [📓] Fine-tuning Phi-4-mini-reasoning with Apple MLX
      • [📓] Phi-4-mini-reasoning with GitHub Models
      • [📓] Phi-4-mini-reasoning with Microsoft Foundry Models

  • Demos

    • Phi-4-mini demos hosted on Hugging Face Spaces
    • Phi-4-multimodal demos hosted on Hugginge Face Spaces

  • Vision Samples

    • Phi-4 Samples
      • [📓] Use Phi-4-multimodal to read images and generate code
    • Phi-3 / 3.5 Samples
      • [📓]Phi-3-vision-Image text to text
      • Phi-3-vision-ONNX
      • [📓]Phi-3-vision CLIP Embedding
      • DEMO: Phi-3 Recycling
      • Phi-3-vision - Visual language assistant - with Phi3-Vision and OpenVINO
      • Phi-3 Vision Nvidia NIM
      • Phi-3 Vision OpenVino
      • [📓]Phi-3.5 Vision multi-frame or multi-image sample
      • Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET
      • Menu based Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET

  • Reasoning-Vision Samples

    • Phi-4-Reasoning-Vision-15B
      • [📓] Using Phi-4-Reasoning-Vision-15B to detect jaywalking
      • [📓] Using Phi-4-Reasoning-Vision-15B to math
      • [📓] Using Phi-4-Reasoning-Vision-15B to detect UI

  • Math Samples

    • Phi-4-Mini-Flash-Reasoning-Instruct Samples Math Demo with Phi-4-Mini-Flash-Reasoning-Instruct

  • Audio Samples

    • Phi-4 Samples
      • [📓] Extracting audio transcripts using Phi-4-multimodal
      • [📓] Phi-4-multimodal Audio Sample
      • [📓] Phi-4-multimodal Speech Translation Sample
      • .NET console application using Phi-4-multimodal Audio to analyze an audio file and generate transcript

  • MOE Samples

    • Phi-3 / 3.5 Samples
      • [📓] Phi-3.5 Mixture of Experts Models (MoEs) Social Media Sample
      • [📓] Building a Retrieval-Augmented Generation (RAG) Pipeline with NVIDIA NIM Phi-3 MOE, Azure AI Search, and LlamaIndex

  • Function Calling Samples

    • Phi-4 Samples 🆕
      • [📓] Using Function Calling With Phi-4-mini
      • [📓] Using Function Calling to create multi-agents With Phi-4-mini
      • [📓] Using Function Calling with Ollama
      • [📓] Using Function Calling with ONNX

  • Multimodal Mixing Samples

    • Phi-4 Samples 🆕
      • [📓] Using Phi-4-multimodal as a Technology journalist
      • .NET console application using Phi-4-multimodal to analyze images

• Fine-tuning Phi Samples

  • Fine-tuning Scenarios
  • Fine-tuning vs RAG
  • Fine-tuning Make Phi-3 become industry expert
  • Fine-tuning Phi-3 with AI Toolkit for VS Code
  • Fine-tuning Phi-3 with Azure Machine Learning Service
  • Fine-tuning Phi-3 with Lora
  • Fine-tuning Phi-3 with QLora
  • Fine-tuning Phi-3 with Microsoft Foundry
  • Fine-tuning Phi-3 with Azure ML CLI/SDK
  • Fine-tuning with Microsoft Olive
  • Fine-tuning with Microsoft Olive Hands-On Lab
  • Fine-tuning Phi-3-vision with Weights and Bias
  • Fine-tuning Phi-3 with Apple MLX Framework
  • Fine-tuning Phi-3-vision (official support)
  • Fine-Tuning Phi-3 wit Kaito AKS , Azure Containers(official Support)
  • Fine-Tuning Phi-3 and 3.5 Vision

• Hands on Lab

  • Explorin di latest models: LLMs, SLMs, local development and more
  • Unlocking NLP Potential: Fine-Tuning wit Microsoft Olive

• Academic Research Papers and Publications

  • Textbooks Na All You Need II: phi-1.5 technical report
  • Phi-3 Technical Report: One Big Language Model Wey Dey inside Your Phone
  • Phi-4 Technical Report
  • Phi-4-Mini Technical Report: Small but Strong Multimodal Language Models Through Mixture-of-LoRAs
  • Optimizing Small Language Models for In-Vehicle Function-Calling
  • (WhyPHI) Fine-Tuning PHI-3 for Multiple-Choice Question Answering: Methodology, Results, and Challenges
  • Phi-4-reasoning Technical Report
  • Phi-4-mini-reasoning Technical Report

Using Phi Models

Phi on Microsoft Foundry

You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware devices. To try Phi by yourself, start by playing wit di models and customize Phi for your own wahala using di Microsoft Foundry Azure AI Model Catalog you fit learn more at Getting Started wit Microsoft Foundry

Playground Each model get im own playground to test am Azure AI Playground.

Phi on GitHub Models

You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware devices. To try Phi by yourself, start by playing wit di model and customize Phi for your own wahala using di GitHub Model Catalog you fit learn more at Getting Started wit GitHub Model Catalog

Playground Each model get im own playground to test di model.

Phi on Hugging Face

You fit still find di model on di Hugging Face

Playground Hugging Chat playground

🎒 Other Courses

Our team dey produce other courses! Check am out:

LangChain

Azure / Edge / MCP / Agents

---

Generative AI Series

---

Core Learning

---

Copilot Series

Responsible AI

Microsoft dey committed to help our customers use our AI products responsibly, share our learnings, and build trust-based partnerships through tools like Transparency Notes and Impact Assessments. Plenty of these resources dey for https://aka.ms/RAI. Microsoft approach to responsible AI na based on our AI principles of fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.

Big natural language, image, and speech models - like di ones wey dem use for dis sample - fit behave in ways wey no too fair, no reliable, or fit offend people, wey fit cause wahala. Abeg check di Azure OpenAI service Transparency note to sabi all di risks and limits. Di recommended way wey fit help reduce dese risks na to put one safety system for your architecture wey fit detect and stop bad behavior. Azure AI Content Safety dey provide one independent protection layer wey fit detect bad content wey users or AI generate for different apps and services. Azure AI Content Safety get text and image APIs wey make you fit detect harmful material. Inside Microsoft Foundry, di Content Safety service go allow you see, explore and try sample code wey fit detect bad content for different type modality. Di following quickstart documentation go guide you how to make requests to di service.

Another tin to consider na di general application performance. For multi-modal and multi-models applications, we mean say di system go work as you and your users expect, including say e no go generate bad outputs. E important to check di performance of your whole application using Performance and Quality and Risk and Safety evaluators. You fit also create and evaluate using custom evaluators.

You fit evaluate your AI application inside your development environment using di Azure AI Evaluation SDK. If you get test dataset or target, your generative AI application generations go get quantitative measurement with built-in evaluators or custom evaluators wey you pick. To start with di azure ai evaluation sdk to evaluate your system, you fit follow di quickstart guide. After you run evaluation, you fit visualize di results in Microsoft Foundry.

Trademarks

Dis project fit get trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos mean say you gats follow Microsoft's Trademark & Brand Guidelines.
If you dey use Microsoft trademarks or logos for changed versions of dis project, e no suppose cause confusion or make person think say Microsoft sponsor am. Any use of third-party trademarks or logos go dey under those third-party policies.

Getting Help

If you jam wahala or get any question about how to build AI apps, join:

If you get product feedback or error while you dey build, visit:

---

Disclaimer:
Dis dokment don translate wit AI translation service Co-op Translator. Even though we dey try make am correct, abeg sabi say automated translation fit get errors or mistakes. Di original dokment wey e dey for im own language na di correct source. For important matter, make person wey sabi translate am well translate am. We no go carry any yawa wey fit happen if pesin no understand well or interpret am wrong because of dis translation.