The 2026 model map
'AI model' gets used for one vague thing. But which one?
The idea inside
There isn't one kind of model. Base, instruct, reasoning, embedding, multimodal, image, voice, and agentic are roles: one frontier assistant can play several at once, and real products wire several together.
After this lesson
You can name the main model families, base, instruct, reasoning, embedding, multimodal, image or video, voice, and agentic, and say what each is for.
Where it leads
That's the toolbox. The rest of the course is how to wire it into something that ships.
Inside this lesson
That's the real lesson stage, paused. Claim your pass to operate it.
See how AI actually works, end to end.
This lesson is one stop on the full arc. Unlock all of it, and keep it for life.
What you get
- The 34-lesson main path, a finishable route from a word to agents
- Goal tracks for using AI at work and building AI features
- Boss labs that make you apply a whole act, not just recognize it
- Spaced recall that brings each idea back before you forget
- Course memory: every term defined, with links to where it first appears
- A shareable capability card when you finish the main path
- Lifetime access on every device, every future lesson included
Not videos to watch. You predict, operate the machine, then prove it. That is why it stays.
99 interactive lessons and challenges. No videos, no code.
Free launch pass: lifetime access, no card needed
New here? The first lessons are free to try. Start with lesson 0.1
What this lesson shows
There isn't one kind of model. Base, instruct, reasoning, embedding, multimodal, image, voice, and agentic are roles: one frontier assistant can play several at once, and real products wire several together.
The question it opens with
'AI model' gets used for one vague thing. But which one?
The walkthrough, in the lesson's own words
- A chat model can't search. That's a different family's job.
- You need to search your own documents. Which model does that job?
- There isn't one 'AI'. Tap a family to see what it's for.
- Three roles, one product. That's how 'AI inside' really looks.
- One product usually wires several of these together. Wire this one.
- You need to find the right passage across 10,000 PDFs. Which model does that?
- Searching meaning is an embedding model's job, it turns text into vectors so close meanings sit near each other (1.2, 5.2). A bigger chat model doesn't help. Different job, different family.
- The map has folded: a frontier assistant is now multimodal by default, tool-native, and reasons on a dial, a built-in dispatcher (the router) decides when to think longer. So read these families as roles one model can play several of, not separate species. Specialists still win where cost, speed, or control matter.
- Every family here comes in two flavors. Open-weight models (the DeepSeek, Qwen, GLM class) are files you can download: run them on your own machines, keep data private, fine-tune them. By mid-2026 they sit within months of the frontier at a fraction of the price, and carry much of the traffic on neutral model routers, the aggregator marketplaces that route your request to whichever open model you pick.
- Closed API models (GPT, Gemini, Claude) hold peak capability, but you rent them through an API and control nothing inside. Buyers weigh it per job: cost, privacy, and control pull toward open weights; maximum capability pulls toward the closed APIs (E.7).
- A support assistant: it hears customer calls, searches your help docs, and drafts replies for your team.
- You can read the map now. When a product says 'AI inside', name the roles: voice in, embedding search, assistant out. One frontier model may even play two of those roles itself; the doc search still needs an embedding model.
Key takeaway
I can tell the model families apart and say which job each one does.
What you can do after this lesson
You can name the main model families, base, instruct, reasoning, embedding, multimodal, image or video, voice, and agentic, and say what each is for.
Check yourself: You need to search thousands of your own documents for the right passage. Which model family?
- An embedding model(correct)
- A bigger chat model
- An image model
- A reasoning model
Searching your own text by meaning is what an embedding model does: it turns text into vectors so close meanings sit near each other. A chat model doesn't search, and a bigger one still doesn't.
Where it leads: That's the toolbox. The rest of the course is how to wire it into something that ships.
This is the written summary. The lesson itself is interactive: you predict, drag and operate the mechanism above, and the reveal answers you.