Number of tokens from predictions that were accepted.
Represents tokens generated through predictive mechanisms that were validated and accepted as part of the final response. Higher accepted prediction rates indicate more efficient generation and better prediction accuracy in the AI processing pipeline.
Number of tokens used for reasoning and analysis.
Indicates how many tokens were consumed during the AI's internal reasoning process, including analysis, planning, and decision-making activities that contribute to generating high-quality responses but are not directly visible in the final output.
Number of tokens from predictions that were rejected.
Represents tokens generated through predictive mechanisms that were subsequently rejected or replaced during the generation process. While these tokens contribute to computational cost, they also indicate the AI's quality control and self-correction mechanisms.
Total amount of output tokens generated.
Represents the complete output token generation including all types of generated content such as reasoning, accepted predictions, and rejected alternatives. This total provides the baseline for output generation cost analysis and quality assessment.
Detailed output token usage statistics with generation type analysis.
Provides comprehensive tracking of output token generation, categorizing tokens by their generation type and purpose. This analysis helps understand the AI's reasoning process, prediction accuracy, and overall efficiency in generating high-quality responses and development artifacts.