Starting June 1, 2026, GitHub Copilot is moving from a request-based system to credit-based billing. Your base subscription price stays the same, but premium features like agent mode, code reviews, and advanced model selection now cost AI Credits based on token usage. If you use Copilot daily, this change will affect how you think about your AI tooling budget.

This is a significant shift. Under the old model, you got a fixed number of premium requests per month. Each request counted the same whether it was a simple autocomplete or a complex agent mode session. Under the new model, every premium action consumes credits proportional to how many tokens it uses. A quick chat question might cost a handful of credits. A multi-file refactor through agent mode could cost hundreds.

Light users probably will not notice a difference. Heavy users, especially those relying on agent mode and AI-powered code reviews, should expect higher bills.

Here is everything you need to know about the new pricing, what counts as credit usage, and how it affects your workflow.

What Changed

GitHub replaced its “premium requests” system with AI Credits. Previously, each Copilot tier included a set number of premium requests per month (for example, 300 for Pro users). Once you hit the limit, you were downgraded to a base model until the next billing cycle. Every request counted equally, whether it was a one-line completion or a complex multi-file edit.

Now, instead of counting requests, GitHub counts tokens. Every interaction with a premium model consumes credits based on the number of input and output tokens processed. More complex prompts and longer responses cost more credits. This is similar to how OpenAI and Anthropic price their APIs, but wrapped in a consumer-friendly credit system rather than raw per-token billing.

The key changes:

• Premium requests are gone. AI Credits replace them entirely.
• Token usage determines cost. A short autocomplete suggestion costs far fewer credits than a multi-file agent mode session.
• Base subscription prices stay the same. You still pay $10/month for Pro, $39/month for Pro+, $19/seat/month for Business, and $39/seat/month for Enterprise.
• Overages are possible. When your included credits run out, you either pay for additional credits or get downgraded to the base model, depending on your plan settings.

This aligns with the broader industry trend of moving away from flat-rate AI subscriptions. Fixed pricing made sense when usage was predictable, but agent mode and code reviews consume far more compute than simple autocomplete.

Pricing Breakdown by Tier

Base prices have not changed, but the credit allocations and overage policies differ by tier.

GitHub has not published exact credit-per-dollar conversion rates yet, but the structure is clear: higher tiers get more credits, and organizations can set spending caps to prevent runaway costs.

Promotional Credits for Business and Enterprise

To ease the transition, GitHub is offering bonus credits for the first three months after the switch:

• Business plans receive an extra $30/month in AI Credits per seat for the first 3 months (June through August 2026).
• Enterprise plans receive an extra $70/month in AI Credits per seat for the first 3 months.

These promotional credits stack on top of the credits already included in your plan. For a 50-person Enterprise team, that is an extra $3,500/month in credits during the promotional window.

This is a smart move. It gives organizations time to understand their actual usage patterns before the promotional period ends and they are billed at standard rates. If you are on a Business or Enterprise plan, use this window to audit how your team consumes credits and set appropriate spending limits. Do not assume your September bill will look like your June bill.

What Costs Credits

Not everything in Copilot uses AI Credits. Standard code completions using the base model remain included in your subscription at no extra credit cost. Credits are consumed when you use premium features:

• Agent mode: Multi-step coding tasks where Copilot plans, writes, and iterates on code across files. This is the most credit-intensive feature because it involves multiple rounds of token-heavy interactions. A single agent mode session can involve dozens of internal API calls as the model reads files, plans changes, writes code, and verifies results.
• Code reviews: AI-powered pull request reviews that analyze diffs, suggest improvements, and flag issues. These process large amounts of context, so they consume significant credits. A review of a 500-line diff will cost considerably more than a review of a 20-line change.
• Premium model selection: Choosing advanced models (like Claude Sonnet, GPT-4.1, or Gemini 2.5 Pro) instead of the default base model costs credits. Different models have different credit costs per token. The most capable models cost the most credits.
• Multi-file edits: When Copilot modifies multiple files in a single operation, the token count adds up quickly. Each file adds context tokens on the input side and generated tokens on the output side.
• Extended context windows: Providing more context (larger files, more repository context) means more input tokens and higher credit consumption.
• Chat with premium models: Even standard chat conversations cost credits when you select a premium model instead of the default.

The general rule: the more tokens in and out, the more credits consumed. A quick inline suggestion costs almost nothing. A 30-minute agent mode session that refactors an entire module could burn through a meaningful chunk of your monthly allocation.

Understanding this distinction is critical for managing costs. You can still use Copilot extensively without touching your credit balance, as long as you stick to the base model for completions and chat. Credits only come into play when you reach for the premium features.

Impact on Different Types of Developers

The shift to usage-based billing creates winners and losers. Your experience will depend entirely on how you use Copilot today.

Light Users: Minimal Change

If you primarily use Copilot for code completions and occasional chat questions, you likely will not notice a difference. The base model remains included in your subscription, and moderate use of premium features should stay within your included credit allocation.

Most developers fall into this category. If you open Copilot chat a few times a day, accept inline suggestions, and occasionally ask it to explain a function, your usage pattern is inexpensive to serve. Your $10/month Pro subscription will continue to feel like a good deal.

Heavy Users: Expect Higher Costs

If you rely heavily on agent mode, run frequent AI code reviews, or consistently use premium models, your effective monthly cost will increase. The old system capped you at a fixed number of premium requests regardless of complexity. The new system charges based on actual compute consumed.

This hits a specific profile of developer hardest: those who adopted agent mode as a core part of their workflow. Running agent mode for an hour a day, having it plan and execute multi-file changes, and iterating on the results generates a lot of tokens. Under the old system, that was 10 to 20 premium requests. Under the new system, it could be a substantial portion of your monthly credits.

Developers who were previously hitting their premium request limits every month should pay close attention to their credit usage in June and July. The real cost of AI coding tools is becoming harder to predict with usage-based models.

Teams and Organizations: New Budget Considerations

Engineering managers now need to think about AI tool spending the way they think about cloud compute costs. Business and Enterprise plans include admin controls for setting spending limits per seat or per team, which helps prevent surprise bills. But it also means someone needs to monitor and manage those limits.

The promotional credits help here. Use the first three months to establish baselines. Track which teams and developers consume the most credits, identify whether that usage correlates with productivity gains, and set informed spending limits before the promotional period ends.

For a broader comparison of how this stacks up against other tools, see our AI coding tools pricing guide.

Timeline

Here is the rollout schedule:

• April 20, 2026: GitHub paused new Copilot signups to prepare for the billing transition. Existing users are unaffected during this period, but no new individual or organization signups are being accepted. Free tier signups are also paused.
• June 1, 2026: Usage-based billing goes live. All tiers switch to the AI Credits system. Promotional credits for Business and Enterprise begin. New signups reopen under the new billing model.
• August 31, 2026: Promotional credit period ends for Business and Enterprise plans. Standard credit allocations apply from September onward. Organizations should have usage baselines established by this point.

If you are considering signing up for Copilot, you will need to wait until the new billing system launches on June 1. GitHub has not announced whether signups will reopen before that date.

The pause on signups is unusual and signals how significant this change is internally. GitHub wants a clean cutover rather than managing two billing systems simultaneously.

The Bigger Picture

GitHub is not doing this in isolation. This follows Anthropic removing Claude Code from the Pro plan, effectively ending unlimited access to their most powerful coding tool at a flat rate. The pattern is clear: AI providers are discovering that flat-rate subscriptions do not work when a small percentage of power users consume a disproportionate share of compute resources.

The economics are straightforward. An agent mode session that runs for 20 minutes, making dozens of API calls across multiple files, costs the provider significantly more than a user who triggers a few autocomplete suggestions per day. Flat-rate pricing subsidized heavy users at the expense of light users (and provider margins).

We have seen this pattern before in other parts of the tech industry. Cloud computing went through the same evolution: from fixed-price hosting to pay-as-you-go compute. API providers moved from unlimited plans to tiered usage-based pricing. AI tools are following the same trajectory, just on a compressed timeline.

Expect more AI tools to follow this path. The era of “unlimited AI for $20/month” is ending. For a deeper analysis of this trend, read our piece on the end of flat-rate AI subscriptions.

This does not mean AI coding tools are becoming unaffordable. For most developers, the base subscription plus included credits will cover normal usage. But if you have built workflows that depend on heavy agent mode usage or constant AI code reviews, it is time to budget for that accordingly. The developers who will benefit most are those who understand their usage patterns and choose the right tier. Check our best AI coding tools for 2026 guide to compare your options across providers.

FAQ

Will my monthly bill increase automatically?

It depends on your usage. If you stay within your included credit allocation, your bill stays the same as before. If you exceed your credits, you will either be charged for overages (if enabled) or downgraded to the base model for the rest of the billing cycle. Business and Enterprise admins can set hard spending caps to prevent unexpected charges. Individual Pro and Pro+ users should check their account settings to confirm whether automatic overage billing is enabled or disabled by default.

Can I still use Copilot for free?

Yes. The Free tier still exists and gives you access to the base model for code completions and limited chat. You will not have access to premium features like agent mode or advanced model selection without upgrading to a paid plan.

How do I track my credit usage?

GitHub is adding a credit usage dashboard to the Copilot settings page. You will be able to see a breakdown of credits consumed by feature (agent mode, code reviews, model selection) and track your usage against your monthly allocation. Organization admins get additional reporting tools to monitor usage across their team. GitHub has indicated the dashboard will include daily and weekly usage trends, so you can spot patterns before you run out of credits mid-cycle.

What happens if I am in the middle of an agent mode session and run out of credits?

GitHub has not fully detailed this scenario yet, but based on the documentation so far, the session will either complete using the base model or pause and prompt you to purchase additional credits. For Business and Enterprise users, the behavior depends on the spending policy your admin has configured. This is one of the details to watch for when the full billing documentation drops closer to June 1.

What to Do Now

If you are a current Copilot user, here is what you should do before June 1:

1. Check your current usage. Look at how many premium requests you use per month. If you are consistently near or at your limit, you are likely a heavy user who will see higher costs.
2. Understand your workflow. Identify which features consume the most resources. Agent mode and code reviews are the biggest credit consumers.
3. Set spending limits. If you are on a Business or Enterprise plan, work with your admin to configure spending caps before the transition.
4. Evaluate alternatives. If the new pricing does not work for your usage pattern, explore other options in our best AI coding tools for 2026 guide.

The transition to usage-based billing is not inherently bad. It aligns costs with value delivered. But it does require more awareness of how you use these tools, and that is a shift worth preparing for.