GitHub adds per-model token detail to Copilot usage reports
Copilot customers can now trace input, output and cached tokens behind AI-credit consumption instead of seeing only the final credit charge.
GitHub announced on August 11 that Copilot usage reports now show the token activity behind AI-credit consumption for each model. The added detail is available to individual Copilot users and to administrators on Copilot Business and Enterprise, giving both groups a clearer route from a credit charge to the model activity that produced it.
What the report now exposes
For every model represented in the report, GitHub says customers can see input tokens, output tokens, cache-read tokens and cache-write tokens alongside the AI credits consumed. The breakdown is delivered through the downloadable AI usage report on the billing settings page.
This changes the level of evidence available for a cost review. GitHub says the previous report showed the AI-credit total without the token detail underneath it. A customer could see the charge but had less information for explaining whether it came from prompts, generated output or cache activity, and less detail for comparing consumption among models.
The new report does not itself reduce usage or choose a cheaper model. It exposes the components that can support those decisions. An individual can investigate a high-cost period, while a Business or Enterprise administrator can use the same fields when explaining charges to finance, engineering or procurement stakeholders.
Why the cache fields matter
Separating cache reads and cache writes from ordinary input and output prevents all token activity from appearing as one undifferentiated total. That distinction can help teams determine whether a model's observed credit consumption is associated with repeated context, newly written cache data, generated responses or direct prompt input.
The report should still be read within GitHub's billing model. Token counts are the underlying activity described by the update, while AI credits remain the consumption unit shown alongside them. GitHub has not supplied an independent cost study or promised that every workload will become cheaper. The practical value is auditability: users can trace more of the path from model use to the credit total and then decide where further investigation is warranted.
What to check next
Customers should verify that the downloaded report contains the new columns for the models and billing period they need to review. The next useful checkpoint will be whether GitHub adds the same detail to other billing views or APIs, or changes how the report maps token categories to AI credits. Those expansions were not part of the August 11 announcement.
Status
Confirmed. GitHub published the availability and audience details in its official changelog. Internal confidence is medium because the operational description comes from a single first-party source and has not been independently tested for this report.
Sources
Update note: Last reviewed 2026-08-12. We will revise this post if GitHub changes the report fields, audience or access path.
Sources
- GitHub - per-model token breakdown — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.