GitHub has changed the economics of Copilot. As of June 1, the old comfort of paying a flat $10 and treating a short chat and a long autonomous coding run as the same kind of usage is gone; Copilot plans now settle premium usage through GitHub AI Credits.
The conversion is deliberately simple: 1 credit equals 1 cent. Each request is priced against the model API cost for input, output and cached tokens, then deducted from the monthly allowance. Pro includes $10, Pro+ includes $39, Business includes $19 per user and Enterprise includes $39 per user, with temporary promotional credits for Business and Enterprise until August. Autocomplete and Next Edit suggestions still remain free.
GitHub CPO Mario Rodriguez framed the change as a sustainability problem: a quick question and a multi-hour autonomous session used to cost the user the same, while GitHub absorbed rising inference costs. That bargain made sense when Copilot was mainly autocomplete; it is much harder to defend when users ask an agent to run for hours and call models repeatedly.
For light users, the monthly allowance may cover ordinary chat and code help. The real reset is for teams treating Copilot as a long-running agent: once credits run out, administrators must either add budget or let usage stop. The broader signal is that AI coding tools are moving away from unlimited subscription logic and toward metered compute.
Sources: GitHub Copilot is moving to usage-based billing (The GitHub Blog); CocoLoop; Updates to GitHub Copilot billing and plans (GitHub Changelog).