Tip: GitHub's Usage Report Now Breaks Out Cache Read vs Cache Write Tokens Per Model

GitHub Changelog·medium signalXBlueskyLinkedInCopy link

Shipped 2026-08-11, the downloadable AI usage report now shows input, output, cache-read, and cache-write tokens per model alongside the AI credits each consumed, available to Copilot Business/Enterprise admins and individual Copilot users from the AI usage page in billing settings. Splitting cache reads from cache writes is the specific number you need to tell whether prompt caching is actually paying for itself — aggregate token counts hide a cache that's being written far more often than it's hit. Download it before tuning your context strategy, not after.