Starting June 1, 2026, GitHub Copilot replaces its “premium request” allowance model with GitHub AI Credits — a token-based system where you pay for exactly how much the AI computes on your behalf. Here’s a clear, developer-friendly breakdown of what changes, what stays the same, and how to prepare.

What’s Actually Changing?

GitHub Copilot is not raising its base plan prices. What is changing is the unit of measurement for how your usage is tracked. Previously, Copilot counted “premium request units” (PRUs) and silently downgraded you to a cheaper model once you hit the ceiling. That fallback experience is now gone.

In the new model, every interaction with the AI is measured in tokens — the raw units of text the model reads and writes. Those tokens are converted into GitHub AI Credits, which draw down from your monthly plan allotment. If you exceed it, you pay for what you use.

Key formula: 1 GitHub AI Credit = $0.01 USD  ·  Your plan’s monthly dollar value = your credit allotment. A $10/month Pro plan = 1,000 credits per month.

Plans at a Glance

Base subscription prices are unchanged. What’s new is that plan dollars now directly translate to AI credit wallets, plus a new flex allotment layer.

Plan	Price	Billing type	Credits included
Free	$0/mo	Individual	Limited credits + 2,000 completions/mo
Pro	$10/mo	Individual	$10 base + flex allotment
Pro+	$39/mo	Individual (Popular)	$39 base + flex allotment
Business	$19/user/mo	Org — pooled	$19 per user, pooled org-wide
Enterprise	$39/user/mo	Org — pooled	$39 per user, 4-level budgets

For paid individual plans (Pro, Pro+), usage is split into two buckets: base credits (matched 1:1 with your subscription price, never change) and a flex allotment — variable additional usage that GitHub adjusts over time as model efficiency evolves.

What Consumes Credits — and What Doesn’t

Feature	Consumes credits?	Intensity
Code completions (ghost text)	No — unlimited on all paid plans	Free — unlimited
Next Edit Suggestions	No — unlimited on all paid plans	Free — unlimited
Copilot Chat	Yes	Medium
Copilot CLI	Yes	Low–Medium
Copilot Spaces / Spark	Yes	Medium
Agent mode (multi-step tasks)	Yes — multiple model calls per task	High
Copilot Cloud Agent	Yes	High
Copilot Code Review (PR)	Yes — AI credits + GitHub Actions minutes	High
Third-party coding agents	Yes	Varies

Code review is now double-billed: Starting June 1, Copilot’s pull request review feature consumes both AI Credits and GitHub Actions minutes. Self-hosted runners are exempt from the Actions charge.

Why Agentic Workflows Change the Math

Copilot’s agentic features — agent mode, Copilot Cloud Agent, and agentic code review — are fundamentally different from a single chat message. What looks like one request may involve dozens of model calls: planning, tool invocations, context loading, iterative refinement, and output generation.

GitHub has been transparent about this: the platform’s evolution into agentic AI is the primary reason for the billing shift. These workflows consume significantly more compute per task than simple autocomplete or a brief chat response.

Note: A complex agentic session working across a large codebase will consume meaningfully more credits than a quick question in chat. Plan accordingly if your workflow relies heavily on agent mode.

Model Choice Matters Now

Under the new system, the AI model you select directly affects cost. Frontier models designed for deep reasoning cost more per token than lightweight models suited to quick tasks. GitHub’s pricing tables list per-token rates for each available model, and the model used for most interactions is visible to you in the interface.

Tip: Switching to a lighter model for routine questions is one of the easiest ways to stretch your monthly credit allotment — without sacrificing output quality on simple tasks.

Tips to Stay in Budget

• Write clear, concise prompts. Vague prompts lead to longer back-and-forth exchanges and more token consumption.
• Avoid pasting large files or full codebases into chat unless needed. Input context is billed as input tokens.
• Use lighter models for routine tasks and reserve frontier models for genuinely complex work.
• Monitor your usage in the Billing Overview page — GitHub provides real-time credit tracking and a preview bill experience.
• For org admins: set per-user budgets. A $0 user budget disables Copilot for that seat entirely.
• Update your IDE, CLI, and Copilot extensions to the latest stable version.
• Review the preview bill (available since early May) in your Billing Overview before June 1 to estimate projected costs.

Migration Timeline

April 27, 2026	GitHub officially announced the transition to usage-based billing.
Early May 2026	Preview bill experience launched in Billing Overview — users can estimate projected costs before the switch.
June 1, 2026	All monthly Pro and Pro+ plans automatically migrate. Business and Enterprise transition to pooled AI Credits. Annual subscribers remain on PRU billing until plan expires.
Post-June 1	Annual plan holders can convert early to monthly usage-based billing for prorated credits. Annual plans will not auto-renew — GitHub will communicate options before each expiry date.

Annual plan holders: Model multipliers for premium request units increase on June 1 for those who stay on annual request-based billing. GitHub recommends converting to a monthly usage-based plan on June 1 to avoid this.

The Bigger Picture

This shift mirrors how cloud infrastructure pricing evolved — from fixed server licenses to pay-per-use compute. As AI becomes a core part of the software development lifecycle, treating it as a metered resource makes economic sense for both providers and users.

For developers who use Copilot mainly for code completions and occasional chat, the practical day-to-day impact will likely be minimal. The impact will be felt most by teams running heavy agentic workflows, frequent PR reviews, and long multi-file chat sessions.

The key mindset shift: Copilot is no longer just a subscription tool — it’s a compute resource. Use it intentionally, and it remains excellent value.

Need help managing Copilot costs? Whether you’re an individual developer or managing Copilot across a large engineering org, our team can help you optimize usage, control costs, and get the most from AI-assisted development. Website: https://ecanarys.com/ Email: devops@ecanarys.com