Understanding Token Usage in Responses

Responses include a usage section reporting how many tokens the prompt and completion consumed.

What does the 'usage' information in an API response tell you?