GitHub Copilot AI credits replaced premium requests for most individual plans on June 1, 2026. Your plan now supplies a monthly credit allowance; chats, agents and other model-backed features draw from it according to the model and the tokens used. Paid plans still include unlimited code completions and next edit suggestions.
Applies to: GitHub Copilot individual plans under usage-based billing. Existing annual Copilot Pro and Pro+ subscriptions that stayed on the old system are covered separately below. Documentation checked: October 10, 2026.
On this page
GitHub Copilot AI credits by plan
GitHub splits the allowance on paid individual plans into base credits and a flex allotment. Base credits are the fixed part of the subscription. Flex is included too, but GitHub says that portion can change as model prices and efficiency change. The base balance is used first; flex applies automatically after that.
| Plan | Monthly price | Base credits | Flex | Total included |
|---|---|---|---|---|
| Copilot Free | $0 | Not stated | Not stated | An allowance, amount not stated |
| Copilot Student | Free for verified students | Not stated | Not stated | An allowance, amount not stated |
| Copilot Pro | $10 | 1,000 | 500 | 1,500 |
| Copilot Pro+ | $39 | 3,900 | 3,100 | 7,000 |
| Copilot Max | $100 | 10,000 | 10,000 | 20,000 |
Those figures come from GitHub’s current individual billing documentation and plan comparison. They are monthly allowances, not a fixed number of prompts.
For organizations, Copilot Business costs $19 per granted seat per month and contributes 1,900 credits per user to a shared pool. Copilot Enterprise costs $39 per granted seat per month and contributes 3,900 credits per user. Organization billing and policy controls differ from individual plans, so the rest of this guide focuses on individual accounts.
What one AI credit represents
GitHub converts model usage into AI credits at a fixed accounting rate:
- 1 AI credit = $0.01 USD
- 1,000 credits = $10 of metered model usage
- The charge for an interaction depends on the model and the input, output and cached tokens involved.
That makes the new system different from the old premium-request counter. A short question to a lightweight model can cost less than one credit. A long agent session that reads many files, calls a frontier model and produces a large answer can use far more. There is no reliable prompts-per-month conversion without knowing the model, context size and output.
The practical unit is spend, not message count. Pro includes 1,500 credits, which GitHub presents as $15 of monthly model usage even though the subscription costs $10. Pro+ includes 7,000 credits, or $70 at the metered rate. Max includes 20,000 credits, or $200.
What uses the allowance
GitHub lists these model-backed features as consumers of AI credits:
- Copilot Chat
- Copilot CLI
- Copilot cloud agent
- Copilot Spaces
- Spark
- third-party coding agents available through Copilot
Longer conversations cost more because the model receives more context. Agentic work can make several model calls while it searches, edits or validates a codebase. Choosing a more expensive model raises the token cost as well.
Two familiar IDE features do not draw from the AI-credit allowance on paid plans: code completions and next edit suggestions. GitHub says both remain unlimited for paid subscribers, subject to service rate limits. Copilot Free separately includes 2,000 completions per month.
If you mostly accept inline completions, the size of the AI-credit allowance may matter less than it does for someone who keeps cloud agents running across large repositories.
When credits reset
Included credits reset at 00:00:00 UTC on the first day of every calendar month. The reset is fixed; it does not move to match the date on which you first subscribed or the day GitHub charges your card.
Unused credits do not roll over. A user who subscribes on the 20th still gets a fresh included allowance at the next first-of-month reset, while the subscription’s billing date may remain on a different day.
| Clock | What it controls |
|---|---|
| First day of the month, 00:00 UTC | Included AI-credit allowance resets |
| Your account billing date | Subscription renewal and invoice timing |
| Additional-usage budget | Maximum extra metered spend you allow |
What happens after the included balance runs out
On an individual plan, GitHub documents three choices:
- Upgrade. Usage already consumed in the current cycle counts against the larger allowance. GitHub charges the difference between plans rather than billing both plans for the same period.
- Set a budget for additional usage. Extra credits draw down that dollar budget at $0.01 per credit. A $10 budget covers 1,000 credits.
- Wait for the monthly reset.
A budget is permission to incur extra charges, not a second free allowance. GitHub also notes that additional usage can be capped. If you rely on agents for deadline work, set the budget before the included balance is exhausted and keep a ceiling you are willing to pay.
For a managed Business or Enterprise seat, the user may instead see an option to ask the administrator for more budget.
How to choose between Pro, Pro+ and Max
Start with the work that actually consumes credits, not the headline number.
Pro is the sensible baseline when Copilot is mainly an editor companion: frequent completions, smaller chats and occasional agent tasks. Its unlimited paid-plan completions do not consume the 1,500-credit pool.
Pro+ fits regular agent work when repository-wide prompts, code review or premium models make the Pro allowance tight. The included pool is more than four times Pro’s, while the monthly price is $39.
Max is for sustained, high-volume use. At $100 a month and 20,000 included credits, it is a capacity tier rather than a default upgrade. Check a full month’s usage before paying for it.
The full AI coding agent plans tracker keeps the current vendor wording beside each plan. If the decision is also about editor workflow rather than only allowance, read GitHub Copilot vs Cursor.
Premium requests are now a legacy exception
Before June 1, 2026, Copilot measured many interactions in premium request units and applied model multipliers. GitHub then moved current plans to token-based usage and AI credits.
Premium requests still matter only for a narrower group: existing annual Copilot Pro or Pro+ subscribers who chose to remain on the old billing model. GitHub’s transition documentation says those subscribers can:
- keep the annual plan and premium-request accounting until it ends, after which the account is downgraded to Copilot Free;
- cancel and receive a prorated refund, then optionally subscribe to a monthly plan;
- move to a monthly paid plan, with prorated value from the remaining annual term.
Model multipliers belong to that legacy request-based system. They do not apply to current AI-credit billing. If your account still shows a premium-request counter, check whether it is attached to an annual legacy plan before comparing your screen with the credit tables above.
A practical monthly check
A five-minute review at the start and middle of the month is enough for most individual users:
- Open Copilot usage in GitHub’s billing and licensing settings.
- Confirm whether the account shows AI credits or a legacy premium-request balance.
- Note which feature is driving spend: chat, CLI, cloud agent or another agent.
- If agent sessions dominate, shorten the context where possible and stop sessions that no longer have a clear task.
- Let auto model selection handle routine work when it is suitable. GitHub currently documents a 10% model-cost discount for paid-plan use of auto selection in supported Copilot surfaces.
- Set an additional-usage budget only if you accept the possible charge.
Update the client too. GitHub warns that older IDE and extension versions can show stale terminology, incorrect model pricing or inaccurate usage information.
