Back to rankings

GitHub Copilot: Per-Model Token Breakdown in Usage Report

GitHub CopilotView original changelog

GitHub Copilot's AI usage report now breaks down token consumption per model, showing input, output, cache-read, and cache-write token counts alongside the AI credits each model consumed. Previously the report only showed credit totals without the underlying token detail, making it hard to explain or reduce a charge. The breakdown is available to admins on Copilot Business and Copilot Enterprise, and to anyone on Copilot for individuals, via the downloadable AI usage report in billing settings.

Key Takeaways

  • The usage report now shows input, output, cache-read, and cache-write tokens for every model, not just a single credit total.
  • The change makes it possible to trace a specific charge back to its token composition, rather than treating credit totals as a black box.
  • Copilot for individuals users get full access, not just Business/Enterprise admins — a broader rollout than many prior usage-visibility features.
  • The report is a downloadable file from the AI usage page in billing settings, not a live in-app dashboard.
  • It complements, not replaces, the per-session token quota GitHub added to web chat one day earlier, on August 10.
  • The feature reflects GitHub's continued build-out of cost transparency tooling since transitioning Copilot to usage-based AI credit billing in June 2026.

Seeing Exactly Where Tokens Go

GitHub Copilot's AI usage report now breaks down token consumption per model, showing input, output, cache-read, and cache-write token counts alongside the AI credits each model consumed. Previously, the report only showed the AI credit total for each model, making it hard to explain why a particular charge landed where it did or find concrete ways to bring costs down.

What's New in the Report

With the breakdown in place, anyone reviewing the report can trace exactly how a model's input tokens, generated output tokens, and cached context add up to its final credit cost — and share that detail with finance or engineering stakeholders who need to justify or forecast Copilot spend.

Who Gets It

The per-model token breakdown is available to administrators on Copilot Business and Copilot Enterprise plans, and to anyone using Copilot for individuals. It's accessed by downloading the AI usage report from the AI usage page inside billing settings.

The addition builds on a string of usage-visibility improvements GitHub has shipped through 2026, including per-session token quotas in web chat and AI credit visibility by billing cycle, continuing GitHub's push to make Copilot's usage-based billing model easier to audit.