Understand Tokens and Long-Chat Costs
Token-based pricing charges for the content a model processes and generates. Your latest message is only part of the input when earlier conversation or attached material is also included. Check input, output and any cache rates separately on the selected line’s pricing page.
What is a token?
A token is a unit used to represent content for a model. It can correspond to a word, part of a word, punctuation or other encoded content. The count depends on the model and material, so a fixed word count is not an exact billable token count.
The model pricing page separates Input price, Output price and Cache hit rates. Some lines also list cache-write charges and different context-size tiers. Read the rate labels and units together.
Why can a long chat cost more?
When conversation history is included in a request, the model processes that context along with your new message. Longer source documents, more included history and longer replies can increase usage. Cache treatment and context handling depend on the selected service.
Start a New Chat when moving to an unrelated task. For a continuing project, prepare a concise brief containing the decisions, facts and constraints that still matter. Keep the full source material separately so you can supply it again when needed.
Check the settled cost
Use More actions → Profile → Usage Records for the corresponding platform and date range. For API calls, the site’s current documentation describes a separate task-status lookup for the settled cost; see the API quickstart.
For a plain input/output rate without tiers or cache adjustments, an estimate is:
(input tokens × input rate + output tokens × output rate) / rate unit
Use 1,000,000 as the rate unit when prices are quoted per million tokens. Apply any listed tier or cache rules separately, and compare the estimate with the final usage record.
Next: Read model pricing.
Last updated