Lesson 3 of 9 · AI Basics
Model differences and harnesses
Every AI company ships a family of models, not just one. And the model is only half of what you're using. Here's how the pieces fit.
Why so many models?
OpenAI, Anthropic, and Google each offer several models at once. The names change constantly, but the pattern doesn't: every family has a fast, cheap model for quick tasks, a balanced everyday model, and a deep reasoning model that takes longer and thinks harder. As of August 2026, that top shelf includes models like OpenAI's GPT-5.6, Anthropic's Claude 5 family, and Google's Gemini 3 line.
Tap a model tier: "Plan me 3 days in Zion National Park"
The everyday default
Day 1: arrive early, ride the shuttle to Angels Landing (permit required, apply the day before). Day 2: hike The Narrows bottom-up, rent water shoes in Springdale. Day 3: easier morning at Emerald Pools, then scenic drive out through the east tunnel.
Illustrative answers. The pattern is real: bigger reasoning models plan around problems; smaller ones list the obvious.
The practical takeaway: if an answer feels shallow, you may just be on the fast model. Most apps let you switch. Save the deep reasoning models for questions where being right matters more than being quick.
What's a harness?
Here's the part almost nobody explains. The model itself is just a brain in a jar: text in, text out. It cannot browse the web, open your files, or remember last Tuesday. All of that comes from the harness, the app wrapped around the model.
Tap each layer: where your question actually goes
The app around the model: the chat window, saved memory, web search, file uploads, and tools it can run. ChatGPT, the Claude app, and Gemini are all harnesses. So are coding tools like Claude Code. The harness gathers what the model needs and hands it over.
Why harnesses matter
Two apps using the same model can feel completely different, because the harness decides what the model gets to see and do. When ChatGPT searches the web, that's the harness fetching pages and handing them to the model to read. When an AI coding tool edits files on a computer, that's the harness doing the file work while the model decides what to change.
Names you'll hear
A cheat sheet for conversations. Models (the brains): OpenAI's GPT line, Anthropic's Claude models, Google's Gemini models. Chat harnesses (apps for talking): the ChatGPT app, the Claude app, the Gemini app, Microsoft Copilot. Coding harnesses (apps that let AI work on real files and projects): Claude Code, OpenAI Codex, Gemini CLI, GitHub Copilot, and Cursor. Same idea every time: a brain, wearing a harness built for a particular job.
So when you pick an AI tool, you are really picking two things: which brain, and which harness around it. The next lesson compares the big ones you can actually sign up for today.
Last updated August 11, 2026