Choose a task, or enter its size in tokens. A token is a fragment of text, often a short word. The estimate covers the AI model bill only, at three model sizes.
| Model size | Standard | With prompt caching | With caching and an overnight batch |
|---|---|---|---|
| Large | |||
| Mid-size | |||
| Small |
Prices from Anthropic's published list as of October 15, 2026, for Claude Opus 5.5 (large), Sonnet 5.5 (mid-size) and Haiku 5.5 (small), in US dollars. Prompt caching bills the stored part of a request at the cache price; the small charge for first storing it is left out. The batch price is half the standard price.
The estimate leaves out building the system, hosting it, and models that run on your own servers, which carry no fee per item. The right model size is settled by testing on your own cases, as the paper describes.
Your answers stay in your browser.