Skip to main content

AI Usage > Output tokens

Written by Tom Williams

Dataset: AI Usage

Entity: AI Usage Measurement

Field ID: output_tokens

Type: Number

Description: The number of tokens generated by the model.

Source: App

Transformation logic: N/A

App Mapping

GitHub Copilot

token_usage.output_tokens_sum - command line and desktop app records only

Anthropic Claude (coming soon)

tokens.output on Claude Code records; output_tokens on direct API traffic

Cursor (coming soon)

the sum of tokenUsage.outputTokens across the record's usage events

OpenAI Platform (coming soon)

output_tokens

OpenAI Codex (coming soon)

N/A

Reporting Use Cases

The Output tokens field measures how much the model actually produced. It is the most directly comparable consumption measure across the providers that publish tokens at all.

  • Consumption by model: Filter to breakdown = SURFACE_MODEL and group by model_ref for the clearest picture of where model capacity is going.

  • Consumption by person or team: Group by actor_username over aggregation_level = USER records to find the heaviest consumers.

  • Output per interaction: Dividing output tokens by interactions shows how much the model is producing per prompt, which rises sharply as teams move to agentic work.

  • Not available for Copilot IDE work: An empty value on Copilot IDE records means not reported.

Did this answer your question?