GitHub Copilot Premium Requests vs AI Credits: Limits, Resets and Extra Charges
The quick answer
Before upgrading Copilot because your allowance ran out, check which billing system your account uses. Advice about premium-request multipliers will not explain a credit-based bill.
Build your budget shortlistIn this guide
Before upgrading Copilot because your allowance ran out, check which billing system your account uses. Advice about premium-request multipliers will not explain a credit-based bill. Identify what your account measures and which tasks consume the allowance before paying for more capacity.
Which billing system applies to you?
GitHub introduced usage-based billing on June 1, 2026. Some existing annual Pro and Pro+ subscribers who kept legacy billing still use premium requests until their annual term ends. GitHub's billing transition and annual-plan exception explains that distinction.
| What your account uses | How to interpret it | What to check first |
|---|---|---|
| Legacy premium requests | Request units with model multipliers | Whether your existing annual plan remains on legacy billing, and when it ends |
| GitHub AI Credits | Model and token costs expressed in credits | Your current allowance, usage and additional-spending settings |
Read the unit displayed in your own billing view before following an older tutorial or another person's screenshot.
If you kept an eligible legacy annual plan, GitHub says it continues until the term ends, then downgrades to Free. Review the documented refund or upgrade options before changing it; do not assume a switch will preserve the remaining allowance or automatically renew an equivalent paid plan. Legacy annual-plan options
What an AI credit buys
GitHub converts model-priced input, output and cached-token usage into AI credits at 1 credit = USD 0.01. Paid plans include base credits followed by a variable flex allotment. Paid code completions and next-edit suggestions are outside AI-credit billing. Individual billing rules
For example, a hypothetical USD 10 additional-usage budget represents 1,000 credits. The number of prompts it covers depends on the model and token consumption. Keep this extra budget separate from your subscription fee and included allowance.
Keep the units attached when comparing offers. Our guide to comparing usage credits explains why request counts, tokens and dollar-valued usage need different comparisons.
Your reset date can differ from your invoice date
Included AI credits expire rather than rolling over. They reset at 00:00 UTC on the first calendar day of the month, independently of the subscription billing date. Credit reset rules
Eligible legacy premium-request counters also reset at 00:00 UTC on day one. GitHub documents viewing those counters in supported IDEs and billing settings, but that guide applies specifically to legacy accounts. Legacy usage monitoring
When planning work around a deadline, record the displayed reset time separately from your payment date.
Decide whether extra spending should stop
For personal accounts, open GitHub billing, then Budgets and alerts. When creating a budget, choose its product or SKU and scope carefully. A Copilot AI credits budget is narrower than a bundled AI-credits budget that can cover other products.
An email alert alone does not stop spending. Select Stop usage when budget limit is reached, where available, if you want blocking. GitHub's separate user-level budgets enforce hard stops without that checkbox; those administrator-managed controls should not be confused with a personal alert setting. Budget setup and enforcement
Read the saved budget's scope and stopping behavior before starting another long task.
Check one task before buying more capacity
Try a recurring job, such as adding a test for an existing bug. Define the files involved and the check that will establish success. Record the model and usage before and after, then note whether you accepted the result. This proposed check helps you evaluate your own workflow.
GitHub recommends choosing models for the task and limiting context to what the task needs. An agent session can make multiple model calls, so a count of your typed messages will miss work performed during the session. Usage optimization guidance
If the agent repeatedly explores unrelated files or redoes a failed change, fix the task boundary before increasing the budget. Additional allowance can keep that same unproductive session running.
Copilot CLI and SDK offer session limits, which GitHub describes as soft controls for a session. They do not replace overall budgets or spending limits. Check the controls supported by your client rather than assuming an IDE exposes the same setting. Session-limit scope
When does upgrading make sense?
Stay with your current plan if its allowance covers the work you need after removing avoidable retries. For occasional additional usage, compare a bounded extra budget with the cost and conditions of changing plans. If useful, verified work repeatedly exhausts your allowance, compare the next tier using your recorded usage rather than a promised number of prompts per day.
For credit-based individual accounts, GitHub documents waiting for the reset, setting an additional-usage budget, or upgrading as options. Options after included credits run out
Before your next coding session, write down the billing unit, reset time and extra-spending boundary shown in your account. Use that record to decide how much work to assign and when to stop.
Official documentation checked October 19, 2026. This guide covers individual-account billing; organization budgets and student benefits need their own checks.