GitHub Copilot is moving away from a simple-feeling request allowance model and toward something much closer to seat-plus-usage pricing model.
And I for one, am not thrilled about it, because I genuinely enjoyed the simplicity of not having to convert tokens to dollars on every request.
Starting June 1, 2026, GitHub says all Copilot plans will move from premium request units to GitHub AI Credits. Those credits will be consumed based on token usage, including input, output, and cached tokens, using model-specific rates.
That means the bill is going to map more directly to how you actually use Copilot.
For lightweight usage, that may be fine and you may not even notice a change.
For agent-heavy usage, it could feel very different.
The headline price is staying flat, but the economics are changing
GitHub says the base prices are staying the same:
- Copilot Pro stays at $10 per month
- Copilot Pro+ stays at $39 per month
- Copilot Business stays at $19 per user per month
- Copilot Enterprise stays at $39 per user per month
That sounds reassuring.
But the important detail is that these plans now include matching amounts of GitHub AI Credits rather than a request allowance that feels more abstract.
In other words:
- Pro includes $10 in monthly AI Credits
- Pro+ includes $39 in monthly AI Credits
- Business includes $19 per user in monthly AI Credits
- Enterprise includes $39 per user in monthly AI Credits
That is a different kind of pricing story.
It pushes Copilot closer to a model where you pay for access to the product and then consume a usage budget inside it.
Why GitHub is doing this
GitHub is being pretty direct about the reason.
Copilot is no longer just an autocomplete tool or a lightweight chat assistant. It has become much more agentic, with longer sessions, more multi-step work, bigger context windows, and more use of premium models.
That creates a problem for fixed-feeling pricing.
A short question and a long-running autonomous coding session are not remotely the same cost to serve. Under the old model, those could still look surprisingly similar to the customer.
GitHub is basically saying that stopped making sense.
And honestly, that is believable and I don't disagree.
The warning signs were already there last week, when GitHub paused new sign-ups for some individual Copilot plans, tightened usage limits, and changed model availability. At the time, the company framed those changes as reliability measures driven by heavier agentic usage.
Now the broader pricing reset is here.
Premium requests are out, AI Credits are in
This is the clearest mechanical change.
Right now, a lot of Copilot usage is framed around premium requests. Those requests cover more advanced features and models.
Starting June 1, GitHub says premium request units will be replaced by GitHub AI Credits.
Instead of counting requests, GitHub will charge based on token consumption:
- input tokens
- output tokens
- cached tokens
That sounds more precise, and it is.
It also means developers need a different intuition for cost.
Under the old mental model, you could loosely think in terms of how many premium requests you had left.
Under the new model, cost will be shaped by how much context you send, how much text the model generates, which model you use, and how often you retry or run long workflows.
That is a much more cloud-like pricing pattern.
The practical consequence is simple: behavior matters more now
The new system makes some usage patterns a lot more visible financially.
Things that may increase spend faster include:
- very large prompts
- sending lots of repository context
- long model outputs
- repeated retries
- premium model selection
- long-running agent sessions
- parallelized workflows
This is the part developers and managers should not miss.
Even if your sticker price stays the same, your effective Copilot cost can change a lot depending on how your team actually uses the product.
That is why “the monthly price did not change” is not a very useful summary.
The fallback safety net is going away
One underappreciated detail in GitHub’s announcement is that fallback experiences are being removed.
Today, if a user exhausts premium request capacity, GitHub may fall them back to a lower-cost model so they can keep working.
Under the new system, that goes away.
Instead, continued usage will depend on whether you still have AI Credits available and whether your account or admin policy allows additional spend.
That is a meaningful product change, not just a billing change.
It means budget policy becomes part of the user experience.
Individuals should watch annual-plan details and usage limits
If you are an individual Copilot user, there are a few details worth paying attention to.
First, monthly subscribers on Pro and Pro+ will automatically move to usage-based billing on June 1.
Annual subscribers are more awkward.
GitHub says annual Pro and Pro+ users stay on their current premium-request pricing until their plan expires, but model multipliers increase on June 1. When those annual plans expire, users move to Copilot Free unless they choose a paid monthly plan, though GitHub says they can convert earlier and receive prorated credits.
Second, GitHub already tightened individual usage limits in April.
Those limits are separate from premium request entitlements. In other words, you could still have request allowance left and hit a usage limit anyway. GitHub says Pro+ offers more than 5x the limits of Pro.
That matters because the temporary pain a lot of users felt this month was a preview of the broader issue: agentic usage is making flat-feeling plan boundaries harder to maintain.
Teams should stop treating Copilot as a fixed per-seat cost
This is the most practical shift for engineering leaders.
If you run Copilot across a team or company, the old instinct was to think mostly in terms of seat count.
That is no longer enough.
Seat price still matters, but now you also need to care about usage shape.
- Which developers are doing the heaviest agent work?
- Which teams are leaning on code review and cloud agent features?
- Which models are being used most often?
- Which workflows burn tokens without creating much value?
That is now cost-management work.
GitHub is giving admins more tools for it, including budgets at the enterprise, cost center, and user levels.
That is a strong hint about how GitHub expects serious customers to manage the product from here.
Pooled usage is good news for businesses
One genuinely helpful change for teams is pooled included usage.
GitHub says businesses will be able to pool included credits instead of leaving each user’s unused capacity stranded in isolation.
That is a better fit for real organizations, because AI usage is rarely evenly distributed.
Some developers barely touch advanced features.
Others burn through them quickly.
Pooling should make the included usage more economically useful.
It will not solve every surprise-cost problem, but it is smarter than rigid per-user silos.
Copilot code review has a second meter now
Another detail teams should not ignore is that Copilot code review will consume GitHub Actions minutes in addition to GitHub AI Credits.
This matters because it introduces a second billable resource into the workflow.
If your team starts using more AI-assisted review or cloud-agent behavior, the total cost story may span more than one GitHub meter.
That makes monitoring more important, especially for organizations already sensitive to Actions usage.
What developers and teams should do before June 1
A sensible response looks pretty boring, which usually means it is the right one.
- Use GitHub’s preview bill experience as soon as it appears in May.
- Look at which models and workflows are actually driving usage.
- Set budgets before the billing switch, not after the first surprise.
- Identify heavy users and decide whether their usage is high-value or just noisy.
- Watch workflows that combine AI usage with GitHub Actions minutes.
- Teach developers that prompt size, retries, and parallel agent work now affect cost more directly.
- If you are on an annual individual plan, read the transition details carefully.
This is less about panic and more about updating your mental model.
The real story
GitHub is not just changing a billing screen.
It is changing how Copilot’s value and cost line up.
The old premium-request model was a rough abstraction. The new AI Credit model ties usage much more closely to real compute consumption.
That will probably be fairer in some situations.
It will also be less forgiving of vague usage habits.
So the practical takeaway is simple.
Do not think of Copilot as just a flat monthly subscription anymore.
Think of it as a seat-based product with usage economics inside it.
That is the part worth paying attention to.