GitHub Copilot Token Billing: What the June 2026 Change Means for Engineering Teams

GitHub Copilot’s billing model changes on June 1, 2026. The shift from premium request counts to token-based billing — “GitHub AI Credits” — is more than a pricing repackaging. It resolves a fundamental cost inequality that has existed since agentic coding sessions became common.

The problem with request-based billing in an agentic context is that requests are not comparable units. A short chat completion and a multi-step autonomous coding session that refactors a module, runs tests, and opens a PR both counted as “premium requests” under the prior model. The engineering cost of the second is orders of magnitude higher. GitHub’s Chief Product Officer Mario Rodriguez made the internal logic explicit: “Today, a short chat question can cost the user just as much as an autonomous coding session lasting several hours.” Token-based billing aligns what users pay with what compute they actually consume.

The new “GitHub AI Credits” system tracks three token categories: input tokens, output tokens, and cached tokens. Pricing aligns with each underlying model’s API rates. Each subscription tier — Copilot Pro at $10/month, Pro+ at $39/month, Business at $19/user/month, Enterprise at $39/user/month — includes credits matching its monthly cost. Code completions (tab-complete suggestions) remain free and don’t consume credits, preserving the core productivity feature that drove initial Copilot adoption.

The practical cost implications depend heavily on usage pattern. Teams using Copilot primarily for code completions and short chat interactions will see minimal change — their usage profile is token-light. Teams that have adopted agentic coding workflows — running Copilot as an autonomous agent for longer-running tasks — are the ones the new model is designed to price accurately. An autonomous session that reads multiple files, makes cross-repository changes, and iterates on failing tests will consume tokens across a longer context window than a single completion, and will be billed accordingly.

The Decoder’s analysis notes that Business customers receive additional credits during June through August to support the transition. Preview invoices launch in early May, giving teams a window to observe their actual token consumption before the billing change takes effect — which is the recommended action for any team with significant Copilot usage across multiple seats.

There is a separate policy change worth flagging: GitHub announced it will use interaction data from Free, Pro, and Pro+ users for AI training starting April 24, 2026, unless users opt out. This is independent of the billing change but coincides with it. Enterprise and Business tier users are not subject to this policy under current terms.

For engineering leaders budgeting Copilot costs, the transition requires a model shift: from counting seats and assuming flat per-user cost, to measuring token consumption by workflow type. Teams where developers use Copilot for agentic tasks — long-horizon code generation, multi-file refactors, automated test writing — should audit their actual usage patterns using the May preview invoices before the June cutover to avoid billing surprises.

Photo by Lukas Blazek on Pexels

Source: The Decoder

Content created with AI assistance and reviewed for accuracy.