GitHub Copilot: Transition to Usage-Based Billing with AI Credits and New Max Plan
On June 1, 2026, GitHub Copilot moved to usage-based billing, replacing Premium Request Units (PRUs) with a new unit called GitHub AI Credits, where 1 credit equals $0.01 and usage is metered against per-model API token rates. Every plan now includes a monthly credit allotment, and a new Copilot Max tier ($100/month in credits) is available as an upgrade for existing subscribers. Code completions and Next Edit Suggestions stay unlimited, but all other features (including agentic and chat interactions) now consume credits, prompting significant developer concern over potentially much higher bills for heavy agent usage.
Sources & Mentions
5 external resources covering this update
GitHub Copilot is moving to usage-based billing
GitHub Blog
GitHub Copilot is moving to usage-based billing - Discussion #192948
GitHub
GitHub Copilot Switches to Token-Based Billing: Some Developers Fear Costs Will Skyrocket
how2shout
GitHub Copilot's New AI Credits Billing: What Changes June 1
Link
GitHub Copilot Moves to Usage-Based Billing: What Changes June 1
Let's Data Science
GitHub Copilot Moves to Usage-Based Billing
As of June 1, 2026, GitHub Copilot transitioned from its fixed Premium Request Unit (PRU) model to a fully usage-based billing system built on a new currency called GitHub AI Credits. The change fundamentally alters how every interaction with Copilot is measured and charged, and it has drawn substantial reaction from the developer community.
What Replaced Premium Request Units
Under the previous model, each premium model interaction consumed a fixed number of Premium Request Units regardless of how large the request or response actually was. The new system replaces PRUs with AI Credits that are metered against actual token consumption: input tokens, output tokens, and cached tokens are all counted and priced according to the listed API rate for the specific model in use.
The conversion is straightforward: 1 AI Credit = $0.01 USD. Because billing now tracks raw token usage rather than a flat per-request charge, a single long agentic session that reads and writes large amounts of context can consume far more credits than a short chat exchange did under PRUs.
Importantly, GitHub also removed the PRU fallback model. Previously, once a user exhausted their premium allotment, Copilot would continue serving requests on a cheaper, lower-tier model. That fallback no longer exists under the credit system: usage beyond the monthly allotment is billed as overage (on paid plans that opt into additional usage) rather than silently downgraded.
What Stays Free
Not everything is metered. Code completions and Next Edit Suggestions remain unlimited on all plans and do not consume AI Credits. The metering applies to the more compute-intensive surfaces: chat, agent mode, and other premium model interactions.
Plan Allotments and the New Copilot Max Tier
Every Copilot plan now includes a monthly allotment of AI Credits, with paid plans able to purchase additional usage:
- Copilot Pro+ costs $39/month and includes $39 in monthly AI Credits.
- Copilot Business costs $19/user/month with its respective credit allotment.
- Copilot Enterprise costs $39/user/month with its respective credit allotment.
A new top tier, Copilot Max, launched alongside the billing change. Built for heavy Copilot usage, it includes $100/month in AI Credits. As of June 1, 2026, Copilot Max is available only as an upgrade for users who already have an existing Copilot plan.
User-level budget controls, which let individuals cap how much they spend on overage, are now generally available, giving users a guardrail against unexpected bills.
Annual Plans and the Migration Path
GitHub is retiring annual plans. Users currently on an annual subscription continue using PRUs until their subscription expires, after which they move to the credit-based model.
Developer Reaction
The shift triggered a strong response across developer communities. The official GitHub community discussion thread accumulated heavy downvotes, and independent coverage reported that developers running agent-heavy workflows could see effective cost increases ranging from roughly 10x to 50x compared with their previous PRU-based spend, depending on how token-intensive their usage is. The core concern is predictability: token-metered billing makes monthly costs harder to forecast than the old fixed-request model, particularly for teams leaning on long-running agentic sessions.