GitHub Copilot Token Billing Is Live: What Changes

As of today, GitHub Copilot no longer works like a subscription. It works like a cloud meter. If you haven't checked your settings yet, now's a good time.

All GitHub Copilot plans transitioned to usage-based billing on June 1, 2026. Instead of counting premium requests, every plan now includes a monthly allotment of GitHub AI Credits — with the option to purchase additional usage on top.

The plan prices haven't changed. What you get for those prices has.

What's a GitHub AI Credit?

Simple math: 1 AI Credit = $0.01 USD. A $10 budget covers 1,000 credits.

Each plan's monthly credit allowance roughly matches its price. So a $10 Pro plan gets around 1,000 credits. A $39 Pro+ plan gets around 3,900. Business and Enterprise get credits equal to their per-seat cost.

What changed is how those credits get consumed — by token, not by request.

Every interaction now consumes input tokens (what you send), output tokens (what the model generates), and cached tokens (context the model reuses). GitHub prices those tokens per model and converts the total into AI Credits.

Under the old system, a quick one-liner question and a 20-step agentic refactor session cost the same flat premium request. That's over.

What Burns Credits and What Doesn't

This is the most important thing to understand.

Two things stay completely free: inline code completions and Next Edit Suggestions. Everything else that calls a model draws down credits — Copilot Chat, Copilot CLI, agent mode, Copilot Spaces, Spark, and third-party coding agents.

So if you're mainly using Copilot for autocomplete while you type, your bill barely changes. If you're running agentic sessions, multi-file refactors, or long chat threads — that's where you'll feel it.

Copilot code review adds a second meter on top. It now runs on an agentic architecture built on GitHub Actions, so reviewing a pull request with Copilot counts against your included Actions minutes in addition to AI Credits. Your monthly bill can carry two line items from a single workflow.

Monthly vs Annual: Two Different Paths

If you're on a monthly Pro or Pro+ plan, no action needed — you migrated automatically on June 1. Your $10 or $39 becomes a credit allowance of the same value.

If you're on an annual plan, you stay on the old premium request system until your plan expires. But there's a catch: starting June 1, model multipliers increased for annual subscribers still on request-based billing — so the same work now costs more premium requests than before. At expiry you drop to Copilot Free, or you can convert to monthly early and receive prorated credits for remaining value.

Annual plans are being phased out. The long-term path for everyone is monthly usage-based.

What Happens When You Run Out

When your monthly allowance runs out, paid plans can set an additional budget in US dollars to keep working. If you set no budget, usage stops until the next billing cycle.

There's no more quiet fallback to a cheaper model. It just stops. Set a spend cap in your billing settings if you want to avoid mid-sprint surprises.

The good news: you can set a hard cap and make a surprise bill literally impossible. Go do that now if you haven't.

Who Gets Hurt Most

If you use Copilot mostly for autocomplete, you'll barely notice. If any of the following describes you, check your numbers:

• You run agentic coding sessions regularly
• You use Copilot Chat as your main interface
• You use Copilot code review on PRs via GitHub Actions
• You're on a team or enterprise plan with heavy per-seat usage

GitHub paused new sign-ups for Pro, Pro+, and student plans in May, and removed all Gemini models from Copilot on the web. The company's stated reason: agentic usage intensified to the point where "a handful of requests can exceed the plan price." In plain terms, power users were getting more than they paid for, and GitHub is correcting that — across all customers.

It's the end of the comforting fiction that agentic coding could be sold like a gym membership.

Is It Time to Switch?

Worth doing the math. If you're running heavy agentic workflows, the economics are shifting fast. Claude Code doubled its rate limits earlier this year while keeping a flat-rate model — which starts to look a lot more attractive when Copilot's meter is running on every chat message and agent loop.

The right call depends on your workflow. But if you haven't re-evaluated your AI coding tool spend since this change went live, today's a good day to start.

Source: GitHub Blog