GitHub exposes four Copilot token buckets

GitHub’s Copilot usage report now breaks each model’s AI-credit consumption into four token classes: input, output, cache-read, and cache-write tokens. The August 11 changelog says the view is available to Copilot Business and Enterprise admins and to anyone on Copilot for individuals via a downloadable report. Engineering leaders already tracking AI token budgets as an operating resource can now attribute cache behavior and model mix before imposing blunt usage caps.