Models and pricing

Every model below can be chosen for an agent or for a single chat. Prices are what you are charged per million tokens, drawn from your credit balance as the model works — see billing and credits. Cached input is charged at the cache-read price when a prompt is reused, and at the cache-write price the first time it is stored.

The intelligence rating is a coarse guide to how capable a model is, from one brain (Minimal) to five (Frontier): Minimal and Basic models are the cheapest and quickest and suit short, well-defined work; Standard handles most day-to-day tasks; Advanced and Frontier models reason further through long, multi-step work, at a higher price per token. Price alone is not a reliable guide, which is why the rating is shown separately.