Until October, the Copilot budget conversation had one question in it: Business at $19 a seat, or Enterprise at $39? At GitHub Universe, GitHub added a third answer: Copilot included with GitHub Enterprise, with no Copilot seat charge and every request billed as AI usage. Since then, most of the Copilot conversations we have with clients start the same way: should we switch? Here's how the three options actually compare, where the seat plans still make sense, and what to set up before you change anything.
The three ways to pay
All three run on the same currency. Since June 2026, Copilot's premium usage (chat, agent mode, code review, the cloud agent) is metered in AI credits at $0.01 each, charged at each model's token rate. Code completions and next-edit suggestions don't consume credits on any plan (GitHub blog).
- Copilot Business: $19 per user per month, including 1,900 AI credits per user, pooled across the organization. Usage beyond the pool is $0.01 per credit.
- Copilot Enterprise: $39 per user per month, including 3,900 credits per user, pooled the same way, plus priority access to premium models and new features.
- Copilot included with GitHub Enterprise: $0 per seat for users with a GitHub Enterprise Cloud licence, with the same features and models as Copilot Enterprise. There are no included credits: usage is metered from the first request against your enterprise budget. Users in the account without an Enterprise licence can get the same access through Copilot Standalone at $10 per user per month (Copilot plans).
Two constraints matter. An enterprise can't mix the included model with Business or Enterprise seats; it's one or the other. And the included model is a GitHub Enterprise Cloud offer, available through your GitHub or Microsoft account team today, with self-serve rolling out in stages (GitHub changelog).
The math most teams skip
Here's the part that surprises people. A Business seat's $19 buys exactly 1,900 credits, and overage is charged at the same $0.01 per credit. Because credits are pooled, a team on seats pays whichever is higher: the seat total, or what it actually used. On the included model, it pays only what it used. For users who already hold a GitHub Enterprise Cloud licence, the included model is never more expensive at list price.
Take a 50-developer team on Enterprise Cloud:
- Light usage, $12 per developer per month: Business costs $11,400 a year in seats, of which $4,200 of prepaid credits go unused. The included model costs $7,200.
- Heavy usage, $30 per developer per month: both come to $18,000 a year. The seats are used up and the remaining $6,600 is billed as overage either way.
So the included model saves money when usage is below the seat allowance and costs the same above it. The real question isn't which plan is cheaper. It's whether you have the controls to run a metered service.
When seats still make sense
- You need priority access to new models. That's a Copilot Enterprise benefit, and some teams building on the newest models feel the difference.
- You're on GitHub Enterprise Server. The included model isn't available for Server estates.
- Many of your Copilot users don't have a GitHub Enterprise licence. Standalone at $10 plus usage beats a $19 Business seat only for light users. Because you can't mix models, a large unlicensed population can tip the decision back to seats.
- You've negotiated seat discounts that list-price math doesn't capture. Compare your actual contract, not the pricing page.
- Your finance team needs a fixed number. Seats are a prepaid floor. The included model has no floor, and no ceiling unless you set one.
What really changes: AI becomes a metered utility
Whichever model you choose, Copilot now behaves like cloud spend rather than a licence. Three things drive the bill, and none of them is headcount:
Model choice. Credits are charged at each model's rate, so a team defaulting to a frontier model for every chat costs several times what a team on an efficient model costs. GitHub's new Copilot Fusion routing helps by choosing models per task, but your model policy still sets the boundaries.
Agents. Copilot projects, automations and the cloud agent do far more work per request than chat. They're where the productivity gains are, and also where an unwatched budget goes.
Code review. Copilot code review consumes AI credits, and on private repositories it also uses GitHub Actions minutes. The deeper review levels use more of both.
When a budget runs out, completions and next-edit suggestions keep working, but Copilot Chat, code review and cloud agents stop until there's budget again. That's a reasonable safety net, but you don't want to discover it in the middle of a release week.
A 30-day plan before you switch
- Pull three months of usage. Copilot's usage reports show credits consumed by user and model. That's your real input, not the pricing page.
- Model both options against it. Seats: the seat total or actual usage, whichever is higher. Included: actual usage. Add your Standalone users and any negotiated discounts.
- Set up cost centres and budgets. Budgets work at the enterprise, cost centre and user level. Put alerts on them before the first metered invoice, not after.
- Write a model policy. Decide which models are allowed, and which teams get the expensive ones and for what.
- Decide who runs agents. Start projects and automations with one team, measure what they cost per completed change, and expand on evidence.
- Then talk to your account team. The included model is a sales conversation today. Bring the usage data with you.
For the rest of the GitHub bill (the platform licence, per-committer security and Actions), see our GitHub Enterprise pricing breakdown. For everything else announced alongside this change, read our GitHub Universe 2026 recap.
Talk it through in Toronto
We'll be at GitHub Universe On the Road Toronto on November 18 as a Gold sponsor. Bring your usage numbers and we'll help you work out which model fits.
Deop helps Canadian engineering teams adopt GitHub Copilot with governance and cost controls built in, as the 2026 GitHub Growth Services and Channel Partner of the Year and a Microsoft partner specialized in Agentic DevOps. See our AI and Copilot enablement work or book a discovery.
