Choose an AI

Which AI should you use?

Answer two questions and get one clear pick, where to use it and what it costs. Every company sells a smartest model, a balanced one and a cheap fast one; the right choice depends on the job, not the brand.

Every model

The full lineup

checked · model-watch agent opens a PR on changes
ModelTierBest forIn $/1MOut $/1MContext

Prices are API prices per 1 million tokens (about 750,000 English words). In the apps (claude.ai, ChatGPT, Gemini) you pay a monthly plan instead. * Open-weight models are free to download; you pay for the hardware or a host like OpenRouter.

Benchmarks

What each model is actually good at

Published test scores, side by side. Each column is one skill; the brighter the cell, the higher that model ranks among the others on that test. Hover a score for details, click it for the source. "—" means the test hasn't been run on that model (text-only models, for example, can't take the image test).

Rank the models on one skill

Smarts vs price

Up = smarter overall (Artificial Analysis Intelligence Index). Left = cheaper (blended API price per 1M tokens, log scale). The top-left corner is the best value.

Vocabulary

Words you'll see everywhere

Token
A chunk of text, roughly ¾ of an English word. API pricing counts tokens in and out.
Context window
How much the model can read at once. 1M tokens is several books or a whole codebase.
Thinking / effort
How long the model reasons before answering. More effort: better answers, slower, pricier.
API
Using a model from your own code, paying per use instead of a subscription.
Open-weight
A model you can download and run yourself. Your data never leaves your machine.
Hallucination
When a model confidently makes something up. Check facts, numbers and quotes.