GitHub Copilot switches to usage-based billing today: premium requests are gone, replaced by AI Credits

GitHub Copilot moved all paid plans to usage-based billing on June 1, 2026 — today. The old “premium request” counter is gone, replaced by a unit GitHub calls AI Credits. (Source: The GitHub Blog, 2026-04-27, updated 2026-05-28)

Key facts:

• The change took effect June 1, 2026, on every paid Copilot plan.
• Premium requests are retired. Billing now runs on GitHub AI Credits.
• 1 GitHub AI Credit = $0.01 USD of model usage.
• Credits are metered on actual token consumption — input, output, and cached tokens — priced at each model’s published API rate.
• Code completions and chat on included models stay unlimited and consume zero credits.
• Included credits reset monthly and do not roll over.

What actually changed

Under the old system, one “premium request” could mean wildly different amounts of compute depending on the model and task — GitHub itself said the unit was “hard to predict.” The new model bills you for the tokens you actually consume.

Every paid plan now ships with a fixed dollar amount of included credits, and the plan prices did not change:

Plan	Price	Included AI Credits / month
Copilot Pro	$10 / month	$10
Copilot Pro+	$39 / month	$39
Copilot Business	$19 / user / month	$19
Copilot Enterprise	$39 / user / month	$39

(Source: The GitHub Blog, 2026-05-28)

Once you spend your included credits, you only pay for usage beyond them — and you can cap that with a spending limit or set budgets at the user, team, or org level. Hit the budget and premium usage pauses until the next cycle, so there are no surprise charges.

What this means if you’re building with Copilot

For autocomplete-first users, nothing gets worse: completions and chat on included models are still unlimited. The bill only moves if you lean on agent mode and frontier models, where a single long task now meters every input, output, and cached token at the model’s API rate. Developers reacting to the April announcement summed up the worry bluntly — “you will get less, but pay the same price” (Visual Studio Magazine, 2026-04-27) — because heavy agentic sessions can drain a $10 or $39 credit pool faster than a flat request quota did.

The practical move is the same one we’d give for any metered coding agent: watch your token spend, and know your fallbacks. If Copilot’s agent credits run thin, a local or flat-rate setup like running Claude Code as a daily driver or Cursor Composer 2.5 changes the cost equation. And if you’re choosing which agent to trust with unattended work, our Claude Code vs Codex head-to-head is worth a read before you let credits burn on a run you can’t audit.

What you need to do

If you’re on a monthly Pro or Pro+ plan, you were migrated automatically today. On an annual plan, you keep your existing premium-request allowance until your plan renews, then move to AI Credits. Either way: open your billing dashboard, check the projected-usage estimate GitHub rolled out in early May, and set a spending limit before a frontier-model agent run does it for you.

Sources

• GitHub Copilot is moving to usage-based billing — The GitHub Blog (announced 2026-04-27, updated 2026-05-28)
• GitHub Copilot is moving to usage-based billing — community discussion #192948 — GitHub Community
• Devs Sound Off on Usage-Based Copilot Pricing Change — Visual Studio Magazine, 2026-04-27