↓ Skip to main content
sdd.sh
sdd.sh
  1. Articles/

The Flat-Rate Era Is Over: GitHub Copilot Moves to Token Billing on June 1

·1069 words·6 mins·
Table of Contents

The flat-rate era for AI coding tools is over. GitHub announced that all Copilot plans — Pro, Pro+, Business, and Enterprise — will transition to usage-based billing on June 1, 2026. The new system replaces premium request units (PRUs) with GitHub AI Credits, billed based on token consumption at published API rates.

This isn’t just a Copilot story. It’s the final domino falling: Claude Code dropped tiered plans months ago, Cursor introduced Max Mode credits, and now the last major holdout with a predictable monthly flat rate is joining the consumption economy. If you’re building engineering workflows around fixed AI costs, the math just changed.

What Actually Changes
#

On June 1, every Copilot interaction that touches an AI model starts consuming credits. Here’s the split:

Free (no credits consumed):

  • Code completions
  • Next Edit suggestions

Credit-consuming:

  • Copilot Chat interactions
  • Agentic and multi-step coding sessions
  • Copilot code review (which also now eats GitHub Actions minutes — more on that below)
  • Copilot cloud agent tasks
  • Copilot Spaces

The base plan prices are not changing. Copilot Pro stays $10/month, Pro+ stays $39/month, Business stays $19/user/month, and Enterprise stays $39/user/month. But each plan now includes a monthly AI Credit allotment equal to its price — so Pro gets $10 in credits, Enterprise gets $39.

That’s a significantly smaller budget than the old request-based system provided for heavy users. Pro previously offered 300 premium requests per month; Pro+ offered 1,500. The shift from “N requests” to “N dollars of tokens” makes your budget visible, variable, and potentially exhaustible mid-month.

GitHub is launching a preview bill experience in early May so teams can see projected costs before the switch. Business and Enterprise plans get promotional bonus credits through August ($30 and $70 respectively) to cushion the transition.

The Double-Billing Problem: Code Review and Actions Minutes
#

Buried in the changelog: starting June 1, Copilot code review will also consume GitHub Actions minutes on private repositories. Previously, code review only drew from Copilot PRU allowances. Now it has two meters running simultaneously — AI Credits for the model calls and Actions minutes for execution.

Public repositories are unaffected; Actions minutes remain free there. But for private repos, every review triggered by Copilot becomes a compound cost. GitHub’s recommendation is to audit current Actions usage, verify spending budgets, and brief billing administrators before the transition. That’s enterprise-speak for “this is going to surprise your finance team.”

Developer Reaction: “You Will Get Less, but Pay the Same Price”
#

The community discussion on GitHub has been blunt. The most-shared framing from developers: you will get less, but pay the same price. The concern isn’t the existence of usage-based billing — most engineers understand why AI compute is expensive. The concern is the asymmetry: base subscription prices stay flat while the included value shrinks.

GitHub’s decision also arrives alongside other signals that Microsoft is managing Copilot costs aggressively. New individual and student account signups have been suspended. Anthropic’s Opus models were removed from the $10 Pro plan. The direction is clear: premium capabilities are being ring-fenced behind higher tiers or pay-per-use, and the entry-level experience is narrowing.

Why This Matters Beyond Copilot
#

GitHub Copilot’s billing model shift is significant not because it’s surprising — it was inevitable — but because of what it signals about where the industry is heading.

Agentic sessions are expensive. Code completions are cheap; token-efficient auto-suggestions can run for fractions of a cent per completion. But multi-step agents — the kind that plan a feature, write the code, run the tests, fix the failures, and open a PR — generate thousands of tokens per session. No flat-rate subscription can absorb that at scale. GitHub is doing what Claude Code did months ago: being honest that agents have a different cost structure than copilots.

The $10-a-month AI coding tool is a relic. When Copilot launched at $10/month in 2022, it was primarily an autocomplete product. AI coding tools in 2026 are orchestrators, code reviewers, and autonomous agents. Pricing those workflows like a fixed-rate software subscription was always temporary. Usage-based billing is the correct model for variable, compute-heavy work.

The competitive calculus is shifting. For power users who regularly hit Copilot’s limits, the real-money cost of agentic sessions may push them toward tools with cleaner per-token economics. Claude Code on the Max plan or Cursor Pro+ with credits may end up cheaper for teams running many agent sessions per day. For light users — developers who mostly want code completions and occasional chat — Copilot’s free tier remains compelling.

Code completions staying free is strategic. By keeping the core autocomplete experience outside the credit system, GitHub protects its broadest user base. Copilot’s 15 million+ developer reach was built on the inline suggestion experience. Making that credit-consuming would risk immediate churn to free-tier competitors. The credit system is aimed squarely at agentic use cases where the cost is real.

What to Do Before June 1
#

If your team uses Copilot at Business or Enterprise scale:

  1. Pull your current Actions and Copilot usage data. GitHub’s billing settings show historical consumption. Use it to project what the token equivalent looks like.
  2. Configure spending budgets. GitHub will allow admins to set caps on AI Credit consumption. Set them before June 1 or you may find engineers hitting walls mid-sprint.
  3. Audit which Copilot features you actually use. If your team mostly uses code completions and light chat, the transition is painless. If you’ve been running Copilot Autopilot agents or frequent code reviews on private repos, the cost calculus is different.
  4. Evaluate alternatives with the correct comparison. Compare what $19/user/month in tokens actually buys across Copilot, Claude Code, and Cursor Pro. Token rates and model quality differ. The best tool for your workflow may not be the same after June 1 as it was before.

GitHub’s preview billing tool in early May is exactly the right move to make this transition transparent. Whether the actual numbers turn out to be fair — or confirm the “same price, less value” framing — will become clear very quickly once engineers can see projected spend against real usage.

Sources:

Related