GitHub Copilot's New Usage-Based Billing: What Changed June 1 and What It Costs Now

GitHub Copilot's billing model changed on June 1, 2026. If you are running Copilot across an engineering team, the change from flat-rate to AI Credits billing means your monthly invoice is now variable - and there are new controls to manage that variability. This post breaks down exactly what changed, who should consider Copilot Max, how the double-billing on code review works, and how the new pricing stacks up against Claude Code and Codex.

Last updated: June 10, 2026

Official Sources

Pricing and plan details change frequently. Verify against the official docs above before making purchasing decisions.

What Happened June 1 - Flat-Rate to AI Credits Billing Explained

Before June 1, GitHub Copilot worked on a straightforward seat model: pay a fixed monthly amount per user, get a defined set of features. Simple to budget, simple to explain.

As of June 1, 2026, every Copilot plan now bills based on GitHub AI Credits consumed. Each plan still comes with a monthly included allowance of AI Credits, so for most users using the tool at normal intensity, the change may not produce a higher bill. The shift matters when you go over the included amount.

Here is what the new model means in practice:

• Included credits per plan: each tier (Student, Pro, Pro+, Max, Business, Enterprise) ships with a monthly credit pool.
• Overage billing: once included credits are exhausted, you can continue using Copilot if you have set a spending budget - you will be billed at end of month for those additional credits.
• Individual limits: for individual plans (not org or enterprise), GitHub may limit additional AI Credit consumption based on usage patterns, billing history, and account verification status. If you hit that cap, the path forward is upgrading to the next plan tier.

The practical implication: teams that treated Copilot as a fixed line item now need a spending budget configured, or users will simply stop when their credits run out.

Copilot Max: Who It Is For and What You Get

Copilot Max launched alongside the June 1 billing changes as an upgrade path for existing Student, Pro, and Pro+ subscribers. New user sign-ups remain paused at time of writing; GitHub has signaled they will reopen in the coming weeks.

The Max tier is positioned at power users and intensive workflows. It offers higher included AI Credit limits than Pro or Pro+, and higher spending budget ceilings for additional usage. It also unlocks access to models like Claude Fable 5 that are not available on lower tiers.

If you are regularly running into credit limits on Pro+, or you need to use frontier models like Fable 5 as part of autonomous agent workflows, Max is the intended upgrade path. If you are a casual Copilot user doing standard code completion and occasional chat, the lower tiers are likely sufficient.

For org and enterprise teams, Max also integrates with the new user-level budget controls (covered below), which makes it more tractable to give a subset of power users higher limits without opening up the entire org to uncapped spending.

The Real Cost of Copilot Code Review - Actions Minutes Plus AI Credits

This is the detail most teams will miss. As of June 1, Copilot code review now consumes both GitHub Actions minutes and AI Credits.

If your org is already consuming significant Actions minutes for CI, automated reviews via Copilot will add to that total. The two costs are separate: AI Credits for the model inference, Actions minutes for the compute running the review job.

The default runner for Copilot code review is a standard GitHub-hosted runner. Org admins can now set a default runner at the organization level - this means you can route code review to a self-hosted runner or a custom-configured runner to control the Actions minutes cost. Setting this at the org level applies across all repositories without requiring per-repo configuration.

Before enabling Copilot code review broadly, it is worth auditing your current Actions minutes consumption and estimating what automated review will add. The docs for configuring runners at the org level are at docs.github.com.

User-Level Budget Controls: Practical Setup for Org Admins

User-level budgets are now generally available for organizations and enterprises. This is the primary tool for preventing surprise overages.

Key things admins need to know:

• Universal budget vs per-user overrides: you can set a single budget that applies to all users, then override it for specific users or groups.
• AI Credits budgets are total, not just overage: unlike some spend controls that only gate additional spend after the included pool, user-level budgets for AI Credits control total usage - including the included monthly allowance. This means you can cap a user below their plan's included credits if needed.
• Email notifications: as users approach their budget limits, admins receive email alerts. You can adjust budgets anytime from billing settings - you do not need to wait for the billing cycle to reset.
• Where to configure: billing settings in your organization's GitHub account, under the budget management section.

For larger orgs with a mix of heavy and light Copilot users, the right setup is typically a conservative universal budget, with explicit higher-limit overrides for the engineers who need more headroom. This avoids the situation where a handful of power users consume credits that would otherwise be available to the rest of the team.

Full documentation is at docs.github.com/copilot/concepts/billing/budgets-for-usage-based-billing.

Fable 5 on Copilot: Extra Cost or Included?

Claude Fable 5 became generally available on GitHub Copilot on June 9, 2026. It is the first model in Anthropic's Mythos class, designed for long-horizon autonomous coding and knowledge-work tasks.

A few things distinguish Fable 5 from other models available in Copilot:

Pricing: Fable 5 is billed at provider list pricing under usage-based billing. This is not a flat inclusion in your plan - each use pulls from your AI Credits at Anthropic's published rates. Check the Copilot models and pricing reference for the current credit cost per request.

Data retention requirement: Fable 5 requires data retention. Anthropic retains prompts and outputs for up to 30 days to operate the safety classifiers that detect harmful or abusive use. After 30 days, the data is deleted. According to the GitHub changelog, this retained data is not used to train Anthropic's models.

This is a meaningful departure from the other Claude models in Copilot - Claude Opus 4.8, Sonnet 4.5, and Haiku 4.5 all continue to operate under Zero Data Retention. Enabling the Fable 5 policy constitutes acknowledgement of the data retention requirement.

Availability: Fable 5 is available to Copilot Pro+, Max, Business, and Enterprise users. It requires an admin to explicitly enable the policy in Copilot settings - it is off by default.

Enterprise consideration: for organizations with strict data isolation requirements or compliance postures that prohibit third-party retention of code and prompts, the data retention requirement may be a blocker regardless of performance. For those teams, the other Claude models (under ZDR) remain the safer option until Anthropic's safety architecture evolves to support zero-retention operation of the classifier.

There is no BYOK option for Fable 5 within the Copilot interface - you pay at provider list pricing through AI Credits, or you do not use it within Copilot. If you want Fable 5 with BYOK economics, you would need to use it via the Anthropic API directly, outside of Copilot's interface.

Copilot vs Claude Code vs Codex: Pricing Comparison After the Rebill

The June 1 shift changes the competitive picture for teams deciding between Copilot and alternatives. Here is where the tools stand after the rebill:

The key distinction after the rebill: Claude Code's $20/mo Pro and $100-200/mo Max plans are flat-rate subscriptions. You know what you will pay each month regardless of how many requests you make. Copilot's plans now have a variable component once included credits are exhausted.

For teams that value billing predictability, flat-rate tools have an advantage. For teams that use Copilot lightly and will stay within included credits, the effective cost is unchanged.

For a deeper breakdown across all tools, see our AI coding tools pricing comparison and the Q2 2026 pricing update.

The Privacy Tier Problem

The data retention requirement for Fable 5 surfaces a broader issue with Copilot's privacy model that predates the June 1 billing change. The GitGuardian analysis of Copilot security and privacy outlines the distinction between how different plan tiers handle user data.

For Free and lower-tier individual plans, GitHub has historically used prompts and outputs to improve Copilot models, with opt-out mechanisms. Business and Enterprise plans operate under stricter data isolation - prompts and generated code are not used for model training.

The June 1 billing change does not alter this tier structure. What it does change is the cost model for accessing the higher-tier models within those plans. Teams that were already on Business or Enterprise for data isolation reasons will find the new billing model layered on top of their existing privacy posture - they still have the isolation, but usage costs are now variable.

Fable 5's 30-day retention requirement adds a third state to this picture. It applies only to Fable 5, only when the policy is enabled, and operates within Anthropic's architecture rather than GitHub's. An Enterprise customer who enables Fable 5 has GitHub's enterprise data isolation for the platform, but Anthropic's retention for the specific model's safety operations.

For compliance-sensitive teams, the practical guidance is: treat Fable 5 as requiring explicit legal review before enabling, separate from the general Copilot data handling your org has already assessed.

What Teams Should Do Now

Given the June 1 changes, here is a concrete audit checklist:

1. Check current credit consumption. Review your org's billing settings to see where teams are relative to their included credit pools. The billing dashboard should now show AI Credit usage per plan tier.

2. Set user-level budgets before the next billing cycle. Do not wait. Configure a universal budget that reflects your intended spend ceiling, then override for power users who need higher limits. This prevents surprise end-of-month invoices.

3. Audit Copilot code review enablement. If automated code review is on across repositories, account for Actions minutes in your cost projection. Consider whether routing to a self-hosted runner reduces the marginal cost to acceptable levels.

4. Decide on Fable 5 as a separate policy decision. Do not enable it by default because it is available. Run it past your security and legal teams first given the data retention requirement. If the performance gains are compelling for your autonomous agent workflows, the 30-day retention with no-training guarantees may be acceptable - but make that decision deliberately.

5. Compare against alternatives if flat-rate predictability matters to your team. If your billing and finance teams are pushing back on variable AI spend, tools like Claude Code (flat subscription) or the migrate to Claude Code path may be worth evaluating. The GitHub Copilot guide covers the full feature surface if you want a reference point for what you would be trading away.

6. Set a calendar reminder for when new sign-ups reopen. If you need to add seats to Copilot Max or any of the individual plans, GitHub has signaled sign-ups will reopen in the coming weeks. At time of writing, new user sign-ups for Student, Pro, Pro+, and Max remain paused.

The transition to usage-based billing is not inherently bad for teams - it enables more flexible access to better models and gives admins real spending controls that the flat-rate model never had. The risk is treating it as the same product with the same budget assumptions, and discovering the difference when the invoice arrives.

Frequently Asked Questions

What is GitHub AI Credits billing?

GitHub AI Credits is the usage unit introduced on June 1, 2026 for GitHub Copilot. Each Copilot plan includes a monthly pool of AI Credits. Using Copilot features - completions, chat, code review, agent tasks - consumes credits. Once included credits are exhausted, usage continues if a spending budget is configured and you are billed for additional credits at end of month.

Does Copilot Max have unlimited usage?

No. Copilot Max offers higher included credit limits and higher spending budget ceilings than Pro or Pro+, but it is not unlimited. Once you exhaust included credits, additional usage accrues against your configured spending budget and is billed monthly.

Does Claude Fable 5 in Copilot cost extra?

Yes. Fable 5 is billed at Anthropic's provider list pricing via AI Credits. It is not included in the flat plan price. See the Copilot models and pricing docs for current per-request costs.

Can I use Fable 5 with Zero Data Retention in Copilot?

No. Fable 5 requires 30-day data retention for Anthropic's safety classifiers. ZDR is available for other Claude models in Copilot (Opus 4.8, Sonnet 4.5, Haiku 4.5), but not for Fable 5.

What happens when my Copilot credits run out?

For individual plans, GitHub may limit further usage based on your billing history and account status if no spending budget is set. For org and enterprise plans, usage stops or is gated by admin-configured user-level budgets. Setting a spending budget allows usage to continue with end-of-month billing for additional credits.