There are three components that make up the token pricing:

1. User question/prompt: The initial question sent via the API request. This counts towards your input token usage.
2. System processes: Upon receiving the prompt, Magisterium AI utilizes various mechanisms to intelligently handle the query, including classification, database research, and additional tooling to ensure an accurate answer. These operations involve both input and output tokens, with their usage varying based on the query’s complexity and depth of answer required.
3. AI-generated response: The final output given in the API response, which includes the AI-generated answer, counts towards your output tokens.

The API console usage page allows you to track exactly how many input and output tokens are being consumed by system processes in addition to the more straightforward input and output costs associated with user input and answer generation. The pricing for input and output tokens consumed by system processes is the same as the standard pricing.

Visit the Playground to experiment with tokens consumption.