GitHub Copilot premium requests are the metered unit that determined how much advanced Copilot usage your plan covered, and if you are searching for how they work in mid-2026, you need two answers, not one. First, the mechanics: a premium request is consumed each time you use an advanced Copilot feature, scaled by a per-model multiplier, against a fixed monthly allowance that came with your plan. Second, the news: as of June 1, 2026, GitHub moved Copilot from request-based billing to usage-based billing, and premium requests are now officially labeled "legacy" throughout GitHub's own documentation. Their replacement is GitHub AI Credits, metered at one cent per credit.
Both systems matter today. Annual Copilot Pro and Pro+ subscribers who stayed on their existing plans are still billed in premium requests, and every question about the new credits model (allowances, overages, admin controls) is easier to answer if you understand the system it replaced. Here is the complete picture, with the numbers.
GitHub's definition is simple: a request is any interaction where you ask Copilot to do something, whether that is generating code, answering a question, or reviewing a pull request. Routine interactions, like inline code completions, are unlimited on every paid plan and never touch the meter. Premium requests are the interactions that use more advanced processing, and they draw down a monthly allowance:
The critical nuance is the multiplier. The same question routed to different models can cost 0x, 1x, or 15x your allowance, which is why two developers with identical activity can consume wildly different fractions of the pool.
Under request-based billing, each plan included a fixed number of premium requests per user per month:
Allowances reset on the 1st of each month at 00:00 UTC, and unused requests do not carry over. That reset date is worth knowing because it does not necessarily match your billing cycle. A team that exhausts its pool on the 25th gets fresh allowance on the 1st regardless of when the invoice runs.
Every model available in Copilot carries a multiplier, and your allowance is deducted at that rate. The published tiers, per GitHub's model multiplier reference:
Two modifiers apply on top: auto model selection earns a 10% multiplier discount on paid plans, and Enterprise Cloud requests processed with data residency or FedRAMP enforcement add 10%. And the 13x code review multiplier deserves repeating, because it is the one that surprises teams: an organization that leans on Copilot for pull request review can consume an entire 300-request Business allowance on roughly 23 reviews.
When a user exhausts their allowance, three things can happen. On any paid plan, the included 0x models keep working, slower at peak, but free. If overages are enabled, additional premium requests bill at $0.04 per request at the standard 1x rate (a 3x model interaction costs $0.12, and so on). If overages are not enabled, premium-model access s until the next monthly reset.
Who controls that switch depends on the plan. Individuals set a spending budget in their own billing settings, or upgrade a tier, which was often cheaper than sustained overage: 300 extra requests at $0.04 costs $12, more than the entire Pro plan. In organizations and enterprises, admins must enable the "Premium request paid usage" policy and ensure no budget is set to zero; otherwise developers hit a hard stop mid-month. GitHub's usage reports and per-SKU analytics show consumption both within and beyond the allowance, which is how you identify the small group of heavy users who typically drive most of the burn, a pattern we see across every vendor in seat and usage right-sizing.
On June 1, 2026, GitHub replaced request-based billing with usage-based billing built on GitHub AI Credits: 1 credit = $0.01 USD. Credits are consumed by the same activities that consumed premium requests (chat, agents, code review, CLI, Spark), but metered against the tokens each model actually processes rather than a flat per-prompt multiplier. Completions and next edit suggestions remain unlimited on paid plans and never use credits.
The current allowances:
Note the Business and Enterprise numbers: through September 1, 2026, a promotional window inflates those allowances to 3,000 and 7,000 credits respectively. When it ends, included usage drops by up to 44% overnight with no change to seat prices. We covered the mechanics and the forecasting math in the Copilot credit cliff, and if you administer either plan, that is the deadline to model before it models you.
Who is still on premium requests? Per GitHub's docs, only Copilot Pro and Pro+ subscribers on an existing annual plan who remained on request-based billing after June 1, 2026. Everyone else, including all Business and Enterprise organizations, is on credits now, and the premium-request documentation has moved to a "legacy" section.
The tracking discipline is the same whichever unit you are metered in. Start with what a premium request or credit actually is: our plain-language explainer on premium requests covers the concept, and the budget analysis of the metered model covers why it changed forecasting. Then instrument three numbers: consumption per user (to find the heavy tail), burn rate against the included pool (to know your runway, since the approach in forecasting the credits pool applies directly), and model mix (because the multiplier or per-token rate, not prompt volume, is usually what moves the bill). GitHub's AI usage dashboard and its CSV export are the primary sources for all three, and its budget hierarchy (enterprise, cost center, and per-user budgets with a "stop usage at limit" toggle) is where you enforce the answer.
The harder problem is that Copilot is rarely the whole picture. The assistants engineering teams connect, such as GitHub Copilot, Cursor, Claude, OpenAI's tools, and whatever gets adopted next quarter, each meter in different units on different calendars, and none of them roll up into one forecast. Premium requests, AI credits, Cursor's usage pools, and raw API tokens all have to normalize into one number a CFO can budget: cost per developer, forecast against plan. That cross-vendor method is the subject of our guide to tracking AI code assistant spend across every vendor, and it is the layer Olumia builds for finance teams: read-only connections to the vendors you already use, normalized into forecasts, anomaly alerts, and chargeback.
Premium requests answered "how much advanced Copilot usage does my plan include" with a flat count and a multiplier table; AI credits answer it with a dollar-denominated pool metered on tokens. The constants across both systems: included allowances reset monthly and do not roll over, a small set of heavy users and heavy models drives most consumption, overage policy is a decision someone must make before the pool runs dry, and a September 1, 2026 deadline is about to shrink the included pool for every Business and Enterprise seat. Whether your organization is on the legacy meter or the new one, the bill is now a function of behavior, and behavior is only governable if you can see it before the invoice does.