Learn what this workshop covers and how it will help you build a local AI workflow that fits your own hardware and use cases.

What To Expect

Install LM Studio and run a small local model so you have a working setup before we explore the details.

Basic Install

Learn when local models make sense, where cloud frontier models are stronger, and why local AI skills matter to companies.

Why Choose Local AI?

Introduction

Learn what a model is, how neural networks use weights and parameters, and why open weights are just files full of numbers.

What Is a Model

See how language becomes tokens, how an LLM predicts text one token at a time, and how context and sampling settings affect its output.

What Makes LLMs Special

Learn how CPUs, GPUs, RAM, VRAM, and unified memory affect whether a local model fits and how quickly it can generate text.

How LLMs Run On Hardware

How LLMs Work

Learn how to read a model name and what model families are.

Model Families and Naming

Learn how to search Hugging Face, evaluate a model repository, and identify the right files to download.

Finding Models On Hugging Face

Learn the common types of AI models and how to choose one that matches the task you want to perform.

Model Types By Task

Learn what dense models, mixture-of-experts models, and multi-token prediction mean for local model performance.

Dense Models, MoE, And MTP

Learn why model files need a compatible runtime and how formats such as GGUF and MLX affect a local setup.

Model Formats And Runtimes

The Model Landscape

Learn which hardware details affect local models and how we will use them to find compatible models on Hugging Face.

Matching Models To Your Hardware

See how quantization reduces model memory use, why names such as Q4 and Q8 matter, and how to choose a first download.

Understanding Quantization

Estimate a model's memory needs before downloading it, then verify the estimate with a controlled load test.

Will This Model Fit?

Learn how to read a model card for the prompt template, chat template, and generation settings a local model expects.

Model Configuration

Learn which local-model performance metrics matter and how to compare models without changing every variable at once.

Measuring Local Inference

Hardware, Quantization, And Performance

Learn how LM Studio turns a loaded local model into an OpenAI-compatible API that developer tools can use.

LM Studio API Mode

Learn the shared configuration pattern for connecting an editor, CLI, or coding agent to LM Studio.

Configuring Developer Tools

Connect VS Code to LM Studio in both Agent mode and regular chat mode.

VS Code And Agent Mode

Connecting Pi

Developer Tool Integrations

Learn when remote access makes sense and choose between LM Link and a manual Tailscale connection without exposing your model server publicly.

Remote Access Overview

Use LM Link to make a model loaded on another computer available through your local LM Studio server.

Remote Models With LM Link

Connect a client to an LM Studio server over a private Tailscale network.

Remote Access With Tailscale

Rent a more powerful machine to run your own local model server when buying new hardware does not make sense yet.

Getting Started With Local AI Setup

Getting Started With Local AI Setup

Build the Ultimate Local AI Workflow

Table of Contents

Introduction

How LLMs Work

The Model Landscape

Hardware, Quantization, And Performance

Developer Tool Integrations

Remote Access With LM Studio