Course
BASICModule 2: The AI Landscape· 1/2

A Map of Models

Four big players

In 2026 the market for LLM assistants is held by four ecosystems:

  • Claude (Anthropic) — focused on reasoning quality, working with long documents, and code. Claude Code is the most mature tool for AI automation through the terminal. This course is built around Claude.
  • ChatGPT (OpenAI) — the largest audience, a wide ecosystem of plugins and integrations, strong image generation.
  • Gemini (Google) — deep integration with Google services: Gmail, Docs, Drive, Search. Convenient if you live in the Google ecosystem.
  • Open models (Llama, Mistral, Qwen, DeepSeek) — run on your own hardware, data never leaves. They require technical setup; the quality of top open models is below the flagships, but the gap is narrowing.

Within each ecosystem the models split into tiers: the flagship (maximum quality, more expensive and slower) and lightweight models (faster and cheaper, for simple tasks). At Anthropic this is the Claude family: the larger models for complex reasoning, the smaller ones for high-volume fast operations.

What actually differs

Every vendor's marketing sounds the same; the differences show up in real work:

  • Quality on hard tasks. On simple questions the models have long been indistinguishable. The difference appears on long multi-step tasks: parsing a large document, refactoring code, a chain of reasoning without losing the thread.
  • Handling context. The formal window size is large for everyone; what differs is the ability to actually use details from the middle of a long document.
  • Tools around the model. The chat is a small part of the value. Projects, working with files, running code, agent modes, the API — the ecosystems differ more than the models themselves.
  • Data policy. What happens to your conversations: whether they are used for training, how long they are kept, whether there are no-retention modes. We'll cover this in the safety module.

What you pay for

The subscription scheme is similar everywhere:

  • Free tier — access to mid-range models with tight message limits. Enough to try things out and handle the occasional task.
  • Subscription ($20–25/mo) — access to flagship models, limits several times higher, extra features (Projects, files, search).
  • Top tiers ($100–200/mo) — for intensive daily work: maximum limits, priority access, agent tools. Claude Max is the tier where you can comfortably live inside Claude Code all day.
  • API — pay per token, for scripts and automation. A separate story from the subscription; you'll need it in the pro modules.
💡

A practical strategy: one paid subscription for your main tool plus the free tiers of the rest for comparison on your own tasks. Your own tasks are the only meaningful benchmark: public model rankings measure something other than what you need.

How to choose for yourself

A simple test works: take three of your real tasks (not abstract "write an essay," but your own — with your documents, your code, your specifics) and run them through the candidates. The difference becomes obvious over an evening.

The course bets on Claude for three reasons: quality on long texts and code, the best tooling ecosystem for automation (Claude Code, MCP), predictable behavior on complex instructions. Everything the next modules teach you can carry over to other ecosystems — the principles of prompting and automation are universal.

Key takeaways

  • Four ecosystems: Claude (reasoning, code, automation), ChatGPT (mass reach), Gemini (Google integration), open models (privacy).
  • Models differ on hard tasks and on the tooling ecosystem, not on simple questions.
  • The pricing scheme is similar everywhere: free tier → subscription ~$20 → top tier → API per token.
  • Choose by running your own real tasks, not by public rankings.
CHECK YOURSELF
1. How do you sensibly choose a model for your work?
2. How does a flagship model differ from a lightweight one in the same family?
Hallucinations & LimitsClaude in Detail