Science2 publishers3 min readPublished Updated
GitHub will meter every Copilot plan by token consumption from June 1
Seat prices do not change, but the allowance sitting on top of each seat becomes a dollar balance of tokens, and the fallback to a cheaper model when that allowance runs out goes away on the same day.
The Scientist · Science desk
What happened
- GitHub said every Copilot plan moves to usage-based billing on June 1, 2026, with premium request units replaced by AI Credits drawn down against input, output and cached tokens at each model's API rates.
- Base prices hold at $10 a month for Pro, $39 for Pro+, $19 per user for Business and $39 per user for Enterprise, with Pro and Pro+ carrying monthly credit allotments equal to their price.
- When credits run out there is no longer a drop to a cheaper model, and usage is instead governed by the remaining balance and whatever budget controls an administrator has set.
- Code completions and Next Edit suggestions stay included in every plan and consume no credits, while Copilot code review will bill GitHub Actions minutes in addition to credits.
- Annual Pro and Pro+ subscribers stay on premium-request pricing until their term ends, see model multipliers rise on June 1, and move to Copilot Free at expiry unless they upgrade.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint Team spend now depends on which seats run agent sessions, so a budget built on the average seat under-provisions the few users generating most of the tokens.
- decision With the cheaper-model fallback gone, an administrator has to pick the failure mode in advance: stop the work at a cap, or let the bill run past the included allotment.
- exposure Buyers who planned Copilot as a fixed seat line now carry model price risk on top of the 30% AI infrastructure underestimate IDC forecasts for Global 1,000 companies through 2027.
- contradiction GitHub presents the move as aligning price with usage, while Greyhound Research's Gogia calls that the proximate cause and the breakdown of per-seat pricing under agentic work the structural one.
The only like-for-like conversion the published numbers allow runs through the old overage price. GitHub capped Copilot Pro at 300 premium requests a month from June 2025 and billed anything past that at $0.04 a request [13]. Three hundred requests at that rate come to $12, against the $10 of credits a Pro seat will include after the switch [24]. Enterprise's old cap was 1,000 requests, or $40 at the same rate, against a $39 seat whose allotment InfoWorld reported will match its price [25]. A premium request and a dollar of tokens are different units. The comparison gives a bound and not much more: one unit counted calls, the other counts input, output and cached tokens at each model's listed API rate [2].
Mario Rodriguez, GitHub's chief product officer, wrote that the old unit had stopped tracking cost: "GitHub has absorbed much of the escalating inference cost behind that usage, but the current premium request model is no longer sustainable" [11]. Under premium requests, a quick chat question and a multi-hour autonomous coding session cost the user the same amount [28]. Sanchit Vir Gogia, chief analyst at Greyhound Research, told InfoWorld the sustainability framing was accurate but incomplete, and described the two economic profiles now inside a single seat [16]: "A quiet user nudging completions across a normal working day. A power user orchestrating hour-long edits on a frontier model with heavy context. The first costs almost nothing to serve. The second can cost an order of magnitude more, sometimes considerably more than that" [17].
Usage is calculated on token consumption [2], so the bill tracks how long a session ran and how much context it carried, whether or not the run ended in code anyone merged. Under premium requests an exhausted allowance degraded into a lower-cost model and the work continued, so an overrun surfaced as weaker output; from June 1 it surfaces as a purchase or a halt [6]. GitHub is running the approach with interim brakes. Last week it tightened limits on the Free, Pro, Pro+ and Student plans and paused self-serve Business purchases, calling both reliability and performance measures, and said limits would loosen once usage-based billing is in effect [12].
The pattern predates this announcement. Cursor swapped fixed fast-request allotments for credit pools in June 2025, then apologised publicly and issued refunds after some users incurred large overages [19]. Anthropic bills Claude Code on a token basis through its API with capped subscription tiers layered on top, and OpenAI moved Codex pricing onto token-based credits [20]. Gogia said the convergence across vendors was a workload event being expressed through pricing, not a pricing fashion [22].
The preview bill opens in early May on the Billing Overview page, showing users and admins projected costs before the change [8]. Teams get roughly four weeks of observed spend before June 1 [26].
What to watch
- Whether the early-May preview bill breaks projected spend down by model and by agent session, or shows only one total per seat.
- Whether GitHub loosens the Free, Pro, Pro+ and Student limits and restores self-serve Business purchases once metering is live.
- Whether credits pool across seats for Business and Enterprise customers or expire per user each month.