Foundations

Large Language Models

The brain layer of the AI stack. Billions of weights that generate text by predicting the next token. The big ones (Claude, GPT-5, Gemini) live in the cloud. The open-weight ones (Llama, Qwen, Gemma, Phi, DeepSeek) run on your own machine. This page covers both worlds, plus the Microsoft strategy that sits at the intersection.

In this topic

Everything about large language models

Reference

A complete walk through LLMs and the local-model ecosystem: what a model actually is, open-weight vs closed-weight, the layered AI architecture (model / inference / context / platform), Ollama deep-dive, Microsoft Foundry Local, hardware reality on Mac and Windows, MCP integration, system prompts and why the same model behaves differently across products, and Microsoft's full AI strategy across Phi, Foundry, Copilot, Cowork, and Agent 365.

Last updated May 16, 2026