phi-4

This is a Phi Family of SLMs book for getting started with Phi Models. Phi a family of open sourced AI models developed by Microsoft. Phi models are the most capable and cost-effective small language models (SLMs) available, outperforming models of the same size and next size up across a variety of language, reasoning, coding, and math benchmarks

Microsoft Foundry (demos, documentation, accelerators).

Phi-3, -3.5, and -4 for Mac: Locally-run Vision and Language Models for Apple Silicon

Pivotal Token Search

Notolog Markdown Editor

A secure, on-device document analysis tool using Foundry Local & Microsoft Agent Framework

Voice-first multilingual AI health companion for emergency first-aid and everyday symptom triage in English, Hindi & Urdu built on Microsoft AI Foundry + Phi-4 Reasoning. From emergencies to everyday health, we've got you.

🤖 A simple project consisting of several AIs and the summarization of their responses.___"یک پروژه ساده متشکل از چندین هوش مصنوعی و خلاصه‌سازی پاسخ‌های آنها."

English Transcription + Multi-Layer LLM-Based Subtitle Segmentation & Translation

pRISM is a repository that combines Retrieval-Augmented Generation (RAG) with a multi-LLM voting approach to create accurate and reliable AI-generated outputs. It integrates multiple language models, including Mistral, Claude 3.5, and OpenAI, to enhance performance through advanced consensus techniques

🤖 Microsoft Phi-4 Multimodal Model Demo - A comprehensive Python application showcasing unified text, image, and audio processing with the 14B parameter Phi-4 multimodal model. Features 128K token context, 23+ language support, advanced reasoning capabilities, automated setup, GPU optimization, and real-time multimodal AI processing.

FlowerTune LLM on NLP Dataset

Autonomous cloud security agent — detects misconfigurations, reasons with Phi-4 via Azure AI Foundry, generates Terraform fixes, and opens GitHub PRs automatically.

AWQ Quantization of Microsoft/Phi-4-Reasoning

Enterprise role-readiness agent on Microsoft Foundry. Pick a role, take a mock assessment, get a Foundry IQ grounded learning plan to close your skill gaps.

Local-first Agentic Meeting Copilot. Chrome MV3 tab audio capture, FastAPI/LangGraph backend, faster-whisper + pyannote + phi-4-mini via Ollama, Next.js 15 live dashboard. No paid APIs, no cloud.

Local-LLM agentic pipeline (n8n + Ollama) that turns a raw requirement into validated QA test cases — two independent validation layers (semantic + structural), zero cloud dependency, runs on a 16GB CPU-only laptop. Includes a documented hallucination failure case and honest known-limitations section.

VLM Judges Can Rank but Cannot Score: Task-Dependent Uncertainty in Multimodal Evaluation. Calibrated prediction intervals for VLM-as-a-Judge using only score-token logprobs.

Run large language models locally using Ollama