AnyoneLearnAI
LevelEasyStart
Brain labTry it → read → next · ~9 min

Tutorials · Chapter C (3/4) · ~9 min

Choosing a model

Try it → see it → read → next

Pick for job, cost, privacy, and speed — not hype tweets.

Try yourself

Playground

Model chooser

Turn on what you need. Watch the recommendation reorder — no hype required.

  1. 1
    Swift Mini

    Great for autocomplete & short drafts.

    cheapfast
  2. 2
    Local Open

    Runs on your machine — privacy first.

    privatecheap
  3. 3
    Balanced Pro

    Solid daily driver for most chat tasks.

    smartfast
  4. 4
    Frontier Max

    Hardest reasoning — slower & pricier.

    smart

Best match floats to the top

Recap

What you just did

You practiced picking paths. Model choice is the same muscle.

Teach

How it works

Decision checklist:

  1. Task hardness — classify emails vs write a novel vs prove theorems.
  2. Context needs — long PDFs need larger windows or RAG.
  3. Latency & cost — smaller / faster models for autocomplete UI.
  4. Privacy — on-prem / local vs hosted API; training-on-your-data policies.
  5. Tooling — vision? JSON mode? function calling?
  6. Evals — run your goldens on 2–3 candidates.

Open weights vs closed APIs: control and privacy vs convenience and polish. Both can be right.

Use it

When you'd use this

  • Prototyping: start cheap/fast, upgrade when evals demand.
  • Production: lock versions; don’t silent-auto-upgrade a model under you.

Watch out

Watch out

Version changes break prompts. Pin model IDs. Watch rate limits.

Try next

Try this next

For one personal task, write which model class you’d pick (tiny / mid / frontier) and why.