Which AI model should I use?

The best model for each job

The hard column alone names 4 different models, while everyday requests go to one fast, cheap model, GPT-6 Luna, that answers simple requests as well as the flagships at a fraction of the price. No single model is the right default.

How to choose

• Match the model to the task. Coding, maths, writing and long documents each have a different leader. See the full Router Index for every task and difficulty.
• Don’t pay flagship prices for easy requests. Output prices differ about 100× across these models; a one-line answer doesn’t need the strongest one.
• Let the hard requests think. Reasoning effort matters as much as the model on hard problems, and costs more, so use it only when the request needs it.
• Re-check monthly. New models and price cuts change the winners every few weeks; this page is refreshed from the router each month.

Or don’t choose at all

Nivana makes this choice for every message: it profiles the request, scores all 14 models for it and sends it to the winner, then shows you which one answered and why. Ask it something on the homepage, free and without an account, or start with 2,000 free credits.