On this page · 11 sections
- What GitHub announced on 28 October 2026
- The billing models side by side
- How AI credits are metered
- The core maths: the included credits were never free
- Worked examples at list price
- 16 November 2026: three models leave Copilot
- Which billing model should you pick?
- Settings to fix before a transition takes effect
- India, EU and US notes
- How eCorpIT can help
- References
Summary. On 28 October 2026 GitHub began including Copilot with GitHub Enterprise for enterprises that buy through a GitHub or Microsoft account team on GitHub Enterprise Cloud. Under this model, which GitHub's documentation calls Copilot access, users with a GitHub Enterprise Cloud licence carry a $0 Copilot seat charge, and enterprise members without that licence can take Copilot Standalone at $10 per user per month. Neither option includes AI credits. Chat, CLI, cloud agent and code review are metered from the first request at 1 AI credit = $0.01, while code completions and next edit suggestions stay unlimited. Self-serve customers cannot switch yet, and GitHub has published no dates. Copilot Business ($19, 1,900 pooled credits) and Copilot Enterprise ($39, 3,900 pooled credits) stay on sale, and moving is voluntary. Because each old seat price equals the value of its included credits, an enterprise where every Copilot user holds a GitHub Enterprise Cloud licence pays the same or less at list price, and saves most where the pool goes unused. Standalone users are the exception: once usage passes the old pool value, each one adds $10 a month. Separately, Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra leave Copilot on 16 November 2026.
About this article. This is for engineering, platform and finance leads who pay for GitHub Copilot across an organisation or enterprise. We checked every price, date and rule against GitHub's changelog, GitHub Docs and GitHub's pricing page on 11 November 2026, and recomputed every worked example. The examples use list prices, not negotiated contracts, and GitHub changes Copilot billing often, so re-check before you commit. This is not financial advice, and no vendor paid for inclusion. Researched with AI assistance and reviewed by our editorial team.
What GitHub announced on 28 October 2026
The changelog post, published with GitHub's Universe 2026 announcements, is short; the detail sits in the linked documentation. These are the points that change a budget:
- Copilot is included with GitHub Enterprise for every licensed user in an enterprise that moves to the new model. The docs call this model Copilot access.
- Copilot Standalone costs $10 per user per month for enterprise members who do not hold a GitHub Enterprise Cloud licence. GitHub says the fee pays for identity, access, policy, security and administration controls, not for usage.
- No AI credits are included in either option. Usage is metered from the first request against the enterprise budget.
- The feature set matches Copilot Enterprise. Both options get the same features and models as Copilot Enterprise, subject to admin policies and spending controls.
- Completions never stop. Code completion and next edit suggestions keep working with no budget. Copilot Chat, Copilot code review and Copilot in the cloud need available budget.
- An empty budget never removes the seat. The seat, and the features that need no budget, stay in place.
GitHub's pricing page now lists the Enterprise plan at $21 per user per month, carries a banner saying GitHub Enterprise includes access to Copilot, and names GitHub Copilot access as a plan feature.
What GitHub did not announce matters as much. There is no self-serve date. The eligibility section of the docs says the model is not yet available to customers who manage their own subscription, that GitHub will open it to them in stages and share dates later, and that it is not offered to organisations on GitHub Free or GitHub Team or to self-service Enterprise Cloud trials. Treat any self-serve timeline you hear as unconfirmed until GitHub publishes it.
The billing models side by side
The table below compares the four ways to pay for Copilot in an enterprise, using GitHub's billing comparison and its usage-based billing documentation as read on 11 November 2026.
| Copilot Business | Copilot Enterprise | Copilot access (GHE Cloud licence) | Copilot Standalone (no GHE Cloud licence) | |
|---|---|---|---|---|
| Copilot seat charge | $19 per user per month | $39 per user per month | $0 | $10 per user per month |
| Included AI credits | 1,900 per user per month, pooled | 3,900 per user per month, pooled | None | None |
| Value of included credits at $0.01 | $19 | $39 | $0 | $0 |
| Metered billing starts | After the pool is used | After the pool is used | First request | First request |
| Features and models | Business tier | Enterprise tier | Same as Copilot Enterprise | Same as Copilot Enterprise |
| Who can buy it today | New and existing customers | New and existing customers | Account-team customers on GHE Cloud | Account-team customers on GHE Cloud |
| Unused credits | Forfeited at 00:00 UTC on the 1st | Forfeited at 00:00 UTC on the 1st | Not applicable | Not applicable |
Four rules from the docs shape any transition. First, the billing model applies to the whole enterprise: you cannot run Copilot access alongside Copilot Business or Copilot Enterprise seats in the same enterprise, not even in a single organisation. You can, however, mix licensed users and Standalone users under Copilot access. Second, enterprise owners still assign seats; nobody gets Copilot automatically, including users who hold a GitHub Enterprise Cloud licence. Third, transitions take effect on the first day of a month, agreed with your account team. Fourth, enterprises on Business or Enterprise are never moved automatically.
How AI credits are metered
Some buyers still budget in premium requests. For organisations and enterprises, Copilot usage is now measured in GitHub AI Credits: each interaction's input, output and cached tokens are priced per model, then converted at 1 AI credit = $0.01. GitHub files the older premium-request scheme under request-based billing, marked legacy.
The features that consume credits, per the billing docs, are Copilot Chat, Copilot CLI, Copilot cloud agent, Copilot Spaces, Spark and third-party coding agents. Copilot code review is billed two ways: tokens in AI credits, plus GitHub Actions minutes for the agentic infrastructure that runs the review. The model behind a review is picked automatically and not disclosed, so GitHub publishes estimates instead: an average of about $0.25 per review at the Lite effort level, $1.50 at Balanced and $10 at Max, excluding Actions minutes. Completions and next edit suggestions are not billed in credits and stay unlimited on paid plans.
Two policy settings decide whether metering bills, blocks or runs on unchecked:
- AI credits paid usage. It is on by default. Under Copilot access it must stay on, because with no included allowance, turning it off blocks every feature that consumes credits.
- Stop at the budget limit. The budgets documentation says the setting
Stop usage when budget limit is reachedapplies to enterprise, cost centre and organisation budgets, is off by default, and that without it charges keep accruing past the limit. A budget without this setting is an alert, not a cap.
User-level budgets work differently: they always enforce a hard stop, and a $0 user-level budget blocks credit features at once. The docs also warn that the enterprise budget does not cap licence fees. Their example: 400 Copilot Business licences cost $7,600 a month, so an enforced $5,000 enterprise budget allows up to $12,600 in combined charges. Under Copilot access with no Standalone users, the licence fee for Copilot is $0, so an enforced enterprise limit caps almost the whole Copilot bill.
The core maths: the included credits were never free
The most useful fact in this announcement sits in plain sight. Copilot Business costs $19 and includes 1,900 credits at $0.01 each, which is $19. Copilot Enterprise costs $39 and includes 3,900 credits, which is $39. The old seat price was a prepayment for credits, with a forfeiture rule for anything unused.
Copilot access removes both the prepayment and the forfeiture. For a user who holds a GitHub Enterprise Cloud licence, the monthly comparison at list price is:
Old model (Business): GHE seat + $19 + max(0, usage - $19 pooled share)
Copilot access: GHE seat + $0 + usage
If usage >= pool value: both models cost GHE seat + usage (identical)
If usage < pool value: Copilot access is cheaper by (pool value - usage)
The same holds for Copilot Enterprise with $39 in place of $19. Credits pool across the billing entity, so apply the formula to the enterprise total rather than to individuals.
Standalone users change the result. A Standalone user pays $10 and brings no credits. Once total usage exceeds the old pool value, the new model costs exactly $10 more per Standalone user than the old one. Below that point, the saving from not prepaying shrinks by the Standalone fees.
| Your estate | Usage below old pool value | Usage at or above old pool value |
|---|---|---|
| All Copilot users hold GHE Cloud licences | Copilot access cheaper by the unused pool | Same cost |
| Some users take Standalone | Cheaper until usage reaches pool value minus Standalone fees | Dearer by $10 per Standalone user |
| You are on Copilot Enterprise today | Larger saving, because $39 of prepayment disappears | Same cost, same feature set |
The formula ignores three things: negotiated prices, which GitHub tells you to confirm with your account team; any rise in usage once the visible seat cost drops to $0; and Actions minutes for code review, which bill the same way under both models.
Worked examples at list price
Every example below is illustrative. The seat counts and usage figures are assumptions chosen to show the shape of the maths. The prices are GitHub's list prices on 11 November 2026: GitHub Enterprise at $21 per user per month, and the Copilot prices in the table above.
Worked example 1: 200 developers, all on GitHub Enterprise, Copilot Business today.
- GitHub Enterprise: 200 × $21 = $4,200 a month.
- Copilot Business: 200 × $19 = $3,800 a month, which buys a pool of 200 × 1,900 = 380,000 credits.
At 250,000 credits of monthly use ($2,500), today's bill is $4,200 + $3,800 = $8,000. Under Copilot access it is $4,200 + $2,500 = $6,700, a saving of $1,300 a month. At 520,000 credits ($5,200), today's bill is $4,200 + $3,800 + $1,400 overage (140,000 credits above the pool) = $9,400, and Copilot access is $4,200 + $5,200 = $9,400. Same bill, and every user now has Copilot Enterprise features.
Worked example 2: 100 developers on Copilot Enterprise, light usage.
- GitHub Enterprise: 100 × $21 = $2,100.
- Copilot Enterprise: 100 × $39 = $3,900, a 390,000-credit pool.
If the team uses 120,000 credits ($1,200), today's bill is $2,100 + $3,900 = $6,000, and 270,000 credits are forfeited every month. Under Copilot access it is $2,100 + $1,200 = $3,300, a difference of $2,700 a month. This profile gains most: an enterprise that bought the top tier for features and never came close to its pool.
Worked example 3: 150 licensed developers plus 50 Copilot-only users.
Assume 50 people, such as analysts or contractors whose code lives elsewhere, use Copilot without a GitHub Enterprise licence, and all 200 sit on Copilot Business today.
- Today: GitHub Enterprise 150 × $21 = $3,150, plus Copilot Business 200 × $19 = $3,800 (a 380,000-credit pool).
- Copilot access: GitHub Enterprise $3,150, plus Standalone 50 × $10 = $500, plus metered usage.
| Monthly usage | Credits | Bill today | Bill on Copilot access | Difference |
|---|---|---|---|---|
| $2,000 | 200,000 | $6,950 | $5,650 | $1,300 cheaper |
| $3,000 | 300,000 | $6,950 | $6,650 | $300 cheaper |
| $3,300 | 330,000 | $6,950 | $6,950 | Break-even |
| $4,500 | 450,000 | $7,650 | $8,150 | $500 dearer |
| $6,000 | 600,000 | $9,150 | $9,650 | $500 dearer |
The break-even sits at the old pool value ($3,800) minus the Standalone fees ($500), which is $3,300 or 330,000 credits. Above the old pool, the gap is fixed at the Standalone fees. If your usage export shows this estate running at 450,000 credits, staying on Copilot Business is $500 a month cheaper at list price.
To build the inputs, follow GitHub's own four steps: total your current licence and metered charges; estimate total usage from the AI usage dashboard or export, including usage the pool already covers rather than only overage; count Standalone users and multiply by $10; then compare. Our September credit forecast guide shows how to pull absolute credit numbers rather than percentages, and that method carries over directly.
16 November 2026: three models leave Copilot
GitHub's 15 October changelog deprecates Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra across all Copilot experiences, including Chat, inline edits, ask and agent modes, and code completions, on 16 November 2026. The rates below are per 1 million tokens, from GitHub's models and pricing table on 11 November 2026.
| Model | Status in Copilot | Input | Cached input | Cache write | Output |
|---|---|---|---|---|---|
| Claude Opus 4.8 | Leaves 16 Nov 2026 | $5.00 | $0.50 | $6.25 | $25.00 |
| Claude Opus 5 | Leaves 16 Nov 2026 | $5.00 | $0.50 | $6.25 | $25.00 |
| GPT-5.6 Terra (up to 272K input) | Leaves 16 Nov 2026 | $2.00 | $0.20 | $2.50 | $12.00 |
| Claude Opus 5.5 | Suggested for both Opus models | $4.00 | $0.20 | $5.00 | $20.00 |
| GPT-6.1 Sol (up to 272K input) | Suggested for Terra | $2.00 | $0.10 | $2.50 | $10.00 |
| GPT-6 Luna (up to 272K input) | Suggested for Terra | $0.10 | $0.01 | $0.125 | $0.50 |
| GPT-6.1 Terra (up to 272K input) | Listed, not named as a replacement | $1.00 | $0.05 | $1.25 | $5.00 |
One exception applies: Claude Opus 4.8 stays available to individual Copilot subscribers on annual plans. Business and Enterprise customers lose it.
Under default model enablement, the suggested replacements switch on automatically for Copilot Business and Enterprise unless an administrator has turned off the global default or disabled the model. If your enterprise curates models by hand, enable Opus 5.5 and GPT-6.1 Sol before 16 November, or users will see their preferred model vanish with nothing in its place.
Worked example 4: a month of Opus 5 agent work moved to Opus 5.5. Assume 200 million fresh input tokens, 600 million cached input reads, 20 million output tokens and no cache writes.
- On Claude Opus 5: 200 × $5.00 + 600 × $0.50 + 20 × $25.00 = $1,000 + $300 + $500 = $1,800, or 180,000 credits.
- On Claude Opus 5.5: 200 × $4.00 + 600 × $0.20 + 20 × $20.00 = $800 + $120 + $400 = $1,320, or 132,000 credits.
That is $480 less, about 27%, if token volumes stay the same. They may not: a newer model can use more or fewer tokens for the same task, so measure a fortnight on 5.5 before you lock the saving into a forecast. For trade-offs on coding quality, our best LLM for coding 2027 comparison covers Opus 5.5, GPT-6.1 Sol and Gemini 4 Argon.
Three documentation gaps remain as of 11 November. GitHub's model retirement history table does not yet list the 16 November removals, and both the supported-models list and the pricing table still show all three models as GA. The notice does not mention Claude Opus 4.8 (fast mode) (preview), still priced at $10.00 input and $50.00 output; ask your account team whether it goes too. And GPT-6.1 Terra, now priced at half of GPT-5.6 Terra's input rate, is not among GitHub's suggested replacements, so test it before you treat it as one.
Two other model changes sit in the same window. Gemini 4 Argon and Gemini 4 Carbon carry a promotional Copilot rate of $2.00 input, $0.10 cached input and $10.00 output until 1 December 2026; Carbon then moves to $4.00, $0.20 and $20.00, and GitHub's footnote gives no post-promotion rate for Argon. Carbon arrived on 10 November and, under default model enablement, switches on automatically. Claude Fable 5.2 became generally available on 9 November, billed at provider list pricing, with its admin policy off by default and data retention required by default. Its row is now in the pricing table at $8.00 input, $0.20 cached input, $10.00 cache write and $40.00 output. Our Claude Fable 5.2 pricing note covers the model itself.
Which billing model should you pick?
| Pick this | If |
|---|---|
| Copilot access (GHE Cloud licence) | Every Copilot user already holds a GHE Cloud licence and your usage export shows the pooled credits are not fully used |
| Copilot access with some Standalone users | Your break-even (old pool minus Standalone fees) sits comfortably above your measured usage |
| Stay on Copilot Business | You manage your own subscription (no choice yet), or a large Standalone group pushes usage past break-even |
| Stay on Copilot Enterprise | A negotiated agreement prices your seats below list, so the prepaid pool costs less than metered usage would |
| Wait one billing cycle | You have never exported absolute credit usage, or your spending limits do not stop usage at the limit |
We would push the last row hardest: a $0 seat with an uncapped meter is worse than a $19 seat with a pool if nobody owns the budget.
Settings to fix before a transition takes effect
Transitions land on the first of a month, so work backwards from that date.
- Export a full month of AI credit usage in absolute credits, by user, model and feature. Include usage the pool covered, as GitHub's own estimation steps require.
- Set a universal user-level budget as the baseline for every Copilot-licensed user, then raise individual budgets for measured power users. User-level budgets are always a hard stop.
- Turn on the stop-at-limit setting for the enterprise spending limit and every cost centre and organisation budget. It is off by default.
- Leave AI credits paid usage enabled. Under Copilot access, disabling it blocks Chat, CLI, cloud agent and other credit features entirely.
- Update the firewall allowlist. Copilot access uses the Copilot Enterprise endpoint. If you only allow
*.business.githubcopilot.com, add*.enterprise.githubcopilot.combefore the switch.
- Update clients. GitHub lists minimum versions for accurate billing and usage display, including VS Code 1.120, Visual Studio 2026 18.6.0, the JetBrains plugin 1.9.1 and Copilot CLI 1.0.48. Older versions still work but may show wrong pricing.
- Assign seats deliberately. Copilot access still requires owners to assign seats to members, enterprise teams or organisations. Since 10 November, enterprise GitHub Apps can call the seat management APIs, which helps if you script assignment.
- Re-check model policies for the 16 November removals and for opt-in models such as Claude Fable 5.2.
If your agent and MCP policy is still a wiki page, fix it in the same change window; our guide to Copilot MCP allowlists in enterprise managed settings covers the file and its fail-closed behaviour. The same platform teams also face the GitHub Actions pull_request_target default block on 2 December 2026, which applies to public repositories, with a brownout from 16:00 to 18:00 UTC on 16 November, the same day the models leave. Plan both changes together.
India, EU and US notes
India. Copilot is billed in US dollars, and under Copilot access more of the bill becomes variable. For Indian engineering organisations and global capability centres budgeting in rupees, metered usage moves with both adoption and the exchange rate, and an enforced enterprise spending limit restores a known ceiling. At ₹93.93 per US dollar (ECB reference rate, 10 November 2026), a Standalone seat is ₹939.30 a month, and the $6,700 bill in worked example 1 is about ₹6.29 lakh. GitHub's pricing page lists Enterprise Cloud data residency in the EU and Australia, with more regions promised; India is not listed. Where code or prompts can contain personal data, review which models you enable against the Digital Personal Data Protection Act, 2023. That matters for the Fable family: Fable 5, 5.1 and 5.2 retain prompts and outputs by default for Anthropic's safety classifiers, while other Claude models in Copilot run under zero data retention.
EU. EU data residency is available on Enterprise Cloud, per the pricing page. The Fable retention point applies here too. Eligible enterprises can use the Fable models with zero data retention through the end of 2026 under a time-bound exemption, after which GitHub says continued use will require Anthropic's Enterprise Frontier Safeguards.
US. The pricing page lists FedRAMP Tailored Authority to Operate and annual SOC 1 and SOC 2 Type 2 reports for the Enterprise plan. These are GitHub's published artefacts, not a judgement on your own compliance.
How eCorpIT can help
eCorpIT is a Gurugram-based technology organisation, founded in 2021, assessed at CMMI Level 5 and MSME certified, with senior-led engineering teams working across AWS, Microsoft and Google platforms. Our AI coding agent rollout and governance service models your real Copilot credit usage, runs the Standalone break-even test and sets up budgets with hard stops before a transition date. Talk to us at /contact-us/.
Last updated: 11 November 2026.