Fichas

What Are Tokens?

Before a model can understand text, it breaks the input into smaller pieces called tokens. Tokens are not the same as words - depending on the language and formatting, a single word may consist of one or more tokens.

Both the text you send to the model and the text generated by the model contribute to the total number of tokens processed in a request.

How Tokens Are Used

Every API request consists of three token categories.

Request Breakdown

Example Usage

{
  "usage": {
    "prompt_tokens": 142,
    "completion_tokens": 58,
    "total_tokens": 200
  }
}

Understanding Usage Statistics

Token Lifecycle

Context Window

Every model has a maximum context window, which represents the total number of tokens it can process in a single request.

The context window includes:

• System instructions

• Conversation history

• User messages

• Generated response

Once the limit is reached, older context may need to be removed or summarized.

Estimated Token Consumption

These values are estimates and may vary depending on formatting and language.

Optimizing Token Usage

• Keep prompts concise and focused.

• Remove unnecessary conversation history.

• Summarize older messages instead of sending the full conversation.

• Set an appropriate max_tokens value.

• Choose the right model for your workload.

• Avoid sending duplicate context in every request.

Common Questions

Token Optimization Checklist

• Keep prompts as short as possible.

• Send only relevant conversation history.

• Monitor token usage in API responses.

• Set sensible output limits.

• Review high-usage requests regularly.

Next Step

Now that you understand how tokens influence performance, context, and cost, continue to Responses to learn how NovaAI structures API responses and returns generated content to your application.