GitHub Copilot pricing

About

GitHub Copilot is the AI pair-programmer and coding-agent suite from GitHub (a Microsoft subsidiary). It delivers code completions, an agentic chat experience, a cloud coding agent, automated code review, and a CLI across GitHub.com, VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, Neovim, and more. Launched in 2021 as a completion tool, it has grown into one of the most widely adopted AI developer products and a flagship of Microsoft’s AI strategy.

Copilot serves the full buyer spectrum: individual developers and students on self-serve plans, small and mid-market teams on Copilot Business, and large organizations on Copilot Enterprise (available on GitHub Enterprise Cloud). It competes directly with Cursor, Codeium/Windsurf, Amazon Q Developer, and a growing field of agentic coding tools — but ships with the structural advantage of living inside the world’s largest code host and shipping pull requests, reviews, and Actions where developers already work.

Copilot’s pricing was remarkably static for its first three years — a single $10/month individual seat from late 2021 through 2024 — before changing structure four times in twelve months: a Free tier (December 2024), a Pro+ tier and premium-request metering (April 2025), and finally the dollar-denominated AI Credits model (June 2026). In that last move GitHub re-architected usage billing from a “premium request” quota system to GitHub AI Credits, a pool where 1 credit equals exactly $0.01 USD. That change aligned Copilot’s metered layer with the underlying token economics of the frontier models it routes to (Anthropic Claude, OpenAI GPT-5.x), while keeping bread-and-butter code completions unlimited and free of metering on every paid plan. The migration was the most contentious of Copilot’s history, drawing 767 points on Hacker News and press reports of projected 10×–50× cost increases for heavy agentic users.

Pricing summary : Per-seat subscription plus a dollar-denominated AI Credits pool

GitHub Copilot runs a hybrid pricing model: every plan carries a fixed per-seat subscription, and model-heavy features draw from a monthly GitHub AI Credits allowance where 1 credit equals $0.01 USD. The credit layer is a seat-plus-usage overage mechanism — once included credits run out, additional usage is billed at $0.01 per credit against a US-dollar budget.

• Seat fee — Free $0, Pro $10/mo, Pro+ $39/mo, Max $100/mo (individuals); Business $19/user/mo, Enterprise $39/user/mo (organizations).
• Included AI Credits — each plan bundles a monthly allowance split into Base credits (matched to the subscription price) and a variable Flex allotment. Pro 1,500 · Pro+ 7,000 · Max 20,000 total; Business 1,900/user · Enterprise 3,900/user (pooled).
• What consumes credits — Copilot Chat, agents, cloud agent, code review, CLI, and Spaces, priced by model and token count. Code completions and next-edit suggestions are excluded and stay unlimited.
• Overage — $0.01 per AI credit, capped by user-, cost-center-, and enterprise-level budgets.
• Legacy track — annual Pro/Pro+ subscribers from before June 2026 remain on premium-request billing: 300 (Pro) or 1,500 (Pro+) requests/month, $0.04 per additional request, with per-model multipliers.

What makes this different: GitHub fused a transparent token pass-through ($0.01 = 1 credit) onto a per-seat SaaS plan while carving out completions as an unlimited, non-metered staple — so casual users feel a flat subscription and heavy agent users feel a usage meter.

Pricing by product

GitHub Copilot (Individual plans)

Copilot Student matches Pro features at $0 for verified students (unlimited completions). New Pro/Pro+/Max/student sign-ups have been paused since April 20, 2026; existing customers can still upgrade.

GitHub Copilot (Organization plans)

Promotional period (June 1 – September 1, 2026): existing Business customers receive 3,000 credits/user and existing Enterprise customers 7,000 credits/user before reverting to the standard 1,900 / 3,900. Overage beyond the pool is billed at $0.01 USD per AI credit.

GitHub AI Credits (usage layer)

Legacy premium-request billing (grandfathered annual plans)

Sales motions across products: PLG / self-serve for Free, Pro, Pro+, Max, and Copilot Business; sales-led for Copilot Enterprise and large-volume Business deployments.

Hidden costs : What heavy agent and large-org Copilot usage actually adds up to

The seat fee is only the floor. The real bill is driven by how much agent, chat, and code-review work consumes AI credits beyond the included allowance — and by the 13× multiplier on code review under the legacy request model.

Archetype 1 — A heavy individual agent user on Pro+ ($39/mo). Pro+ includes 7,000 credits (= $70 of underlying model usage). A developer running multiple daily cloud-agent sessions on frontier models can exhaust that and rely on additional budget.

So a single power user can more than double their effective Copilot cost once agent sessions outrun the bundled credits — the meter, not the seat, sets the ceiling.

Archetype 2 — A 50-seat Copilot Business org with active code review. Fifty Business seats pool 95,000 AI credits (1,900 × 50). A team leaning on cloud agents and automated PR review across many repos can burn the pool and spill into per-credit overage.

At moderate agent and review intensity the metered overage can approach two-thirds of the seat spend — which is exactly why GitHub ships user-, cost-center-, and enterprise-level budget caps.

Want to estimate your own GitHub Copilot bill? Use the GitHub Copilot pricing calculator to model your monthly cost based on seats, included credits, and expected agent/chat usage.

Pricing evolution : From premium-request quotas to dollar-denominated AI Credits

For three years Copilot pricing barely moved — a single $10 individual seat from late 2021 through 2024. Then in twelve months it went through four structural shifts: a free tier, a Pro+ tier, premium-request metering, and finally the dollar-denominated AI Credits model. The chronology below is reconstructed from archived github.com/features/copilot pricing pages.

Cadence

Tracked range: 2021-12–2026-06. Quarters not listed were verified stable (0 price changes, 0 SKU additions) against archived pricing snapshots.

Notable changes

• 2021-06-29 — Copilot launched in free technical preview (2,905 points / 1,272 comments on Hacker News), immediately drawing a public-code-training controversy (“GitHub Copilot as open source code laundering?”, 1,028 points).
• 2021-12 — General availability at $10/mo / $100/yr for individuals; this single tier persisted unchanged through 2024.
• 2024-02 — Copilot Business ($19/user/mo) and Copilot Enterprise ($39/user/mo, GitHub Enterprise Cloud only) became generally available.
• 2024-12-18 — A $0 Free plan launched and Copilot Individual was renamed Copilot Pro (still $10/mo); the Free/Pro/Business/Enterprise grid is visible in archived January 2025 snapshots.
• 2025-04-04 — Copilot Pro+ ($39/mo) launched and premium-request metering appeared: Free 50, Pro 300, Pro+ 1,500 monthly premium requests at $0.04 per additional request, with per-model multipliers.
• 2025-06-18 — Enforcement of premium-request limits and paid overage began on GitHub.com (August 1, 2025 on GHE.com), after the enforcement date slipped from an originally-planned June 4.
• 2026-04-20 — New sign-ups for Pro, Pro+, Max, and student plans were temporarily paused (544 points / 230 comments on Hacker News); existing customers retained upgrade paths.
• 2026-06-01 — GitHub AI Credits (1 credit = $0.01 USD) replaced premium requests; individual plans gained Base + Flex allowances, the Max tier ($100/mo) launched, and organizations gained pooled per-user credits with $0.01/credit overage.

The usage-based-billing migration in detail

The June 2026 switch reframed usage from an opaque “request” count — where a single chat and a multi-file agent session both counted as “one request × a model multiplier” — to a token-priced dollar pass-through. Under AI Credits, every interaction’s token cost (input, output, cached) is priced per model and converted at 1 credit = $0.01 USD.

GitHub’s April 27, 2026 announcement drew the loudest pricing reaction of Copilot’s history — 767 points and 553 comments on Hacker News — because the new model also removed the old safety net: when premium requests ran out, Copilot used to fall back to a lower-cost base model so work could continue; under AI Credits, credit-consuming features simply stop until the budget resets or is topped up. Press coverage reported power users projecting 10×–50× monthly cost increases for heavy agentic workflows (e.g. $50 → $3,000 in worst-case projections) once the fallback was removed. To soften the transition, annual Pro/Pro+ subscribers who were mid-term were left on the legacy request model rather than force-migrated, so two billing systems now run in parallel through the annual cycle.

What’s unique : Unlimited completions, dollar-denominated credits, org-pooled allowances

Completions stay unlimited and unmetered. Unlike token-metered competitors, GitHub carves out code completions and next-edit suggestions from credit billing entirely — they are unlimited on every paid plan. The meter only applies to chat, agents, review, CLI, and Spaces, so the staple workflow never produces a surprise bill.

1 credit = $0.01 USD is a literal pass-through. GitHub priced its usage unit at an exact penny, so customers can reason about model spend in dollars. The “credit” abstraction exists mainly to absorb per-model token-rate differences without exposing raw token prices.

Base + Flex split builds in adjustability. Each allowance splits into Base credits (fixed, matched to the seat price) and a Flex allotment GitHub can re-tune “as the economics of AI evolve.” That gives GitHub a lever to keep margins stable as model prices move, without changing the headline seat price.

Organization credits are pooled, not per-seat-locked. Business and Enterprise credits aggregate at the billing-entity level — 100 Business seats share a 190,000-credit pool — so heavy users draw from light users automatically, with budget controls to govern the draw.

Strengths & weaknesses

Billing UX : Budgets, pooled credits, and granular spend controls

• GitHub AI Credits usage dashboard — shows available allowance and consumed credits across IDE, GitHub.com, and CLI in one view.
• User-level budget (ULB) — caps how many AI credits a single user can consume per cycle across both pool and metered phases; always a hard stop (a $0 budget blocks the user immediately).
• Universal vs individual user-level budgets — a default budget applied to every licensed user, overridable per power user.
• Cost center budgets — cap metered charges for a defined group of users or an organization; only active once the shared pool is exhausted.
• Enterprise budget — caps total metered charges across the entire enterprise after the pool is spent.
• Additional-usage budget (US dollars) — set a dollar cap for overage; credits draw it down at $0.01 each (a $10 budget = 1,000 credits).
• Pooled credit allocation — org allowances aggregate at the billing-entity level; adding seats mid-cycle grows the pool immediately, removing them takes effect next cycle.
• Per-feature SKU attribution — Spark and Copilot cloud agent premium usage are tracked in dedicated SKUs for cost visibility.

Strategic wins : Why unlimited completions and a penny-credit unit landed

1. Carving completions out of the meter removed the biggest adoption blocker

By keeping code completions unlimited and never billing them in credits, GitHub ensured the highest-frequency, lowest-margin interaction never triggers a bill. That makes the seat price feel flat for typical users while still metering the expensive agent workloads — a clean answer to the bill-shock problem that plagues pure token meters.

2. The $0.01 = 1 credit unit made usage legible in dollars

Pricing the credit at exactly a penny turned an abstract usage unit into a transparent dollar pass-through, so customers can map model spend directly to budget. It is a textbook example of choosing a value metric buyers can actually reason about. The shift from request quotas to dollar-denominated credits mirrors the broader entitlement-to-credits billing shift sweeping AI products.

3. Org-pooled credits plus four budget levels gave finance real control

Pooling credits across the billing entity and layering user, cost-center, and enterprise budgets means a CFO can cap Copilot spend without blocking the whole team. That governance story is what lets large orgs adopt usage-based billing without fear of runaway invoices.

4. The Base + Flex split future-proofs margins

Splitting each allowance into a fixed Base and a variable Flex layer lets GitHub absorb swings in model pricing by re-tuning Flex, instead of repricing seats every time Anthropic or OpenAI change token rates.

Areas to improve : Dual billing systems, Flex opacity, and review-multiplier surprises

1. Collapse the dual billing systems faster

Running AI Credits and legacy premium requests in parallel for annual subscribers is confusing and support-heavy. GitHub should publish a clear conversion path and target date so no customer has to learn two metering models at once — a lesson other vendors mid-pricing migration should heed.

2. Make the Flex allotment more predictable

Because Flex is explicitly variable, customers can’t fully forecast included usage. A published change-notice window (e.g. 60 days before any Flex reduction) would preserve trust while keeping the adjustability lever.

3. Surface the code-review multiplier before it bites

The legacy 13× multiplier on code review can drain a request allowance fast for teams that review every PR. An in-product estimate (“this will consume ~13 requests”) at review time would prevent silent burn-down.

4. Restore a self-serve net-new individual path

With new Pro/Pro+/Max sign-ups paused since April 2026, net-new individual developers can’t onboard to paid tiers. Reopening a capped self-serve path would protect the product-led funnel that built Copilot’s adoption.

5. Bring back a graceful degradation path

The biggest driver of the 767-point Hacker News backlash to the April 2026 usage-based-billing migration was that AI Credits removed the old fallback — once premium requests ran out, Copilot used to drop to a cheaper base model so work could continue; now credit-consuming features simply stop. Press reported power users projecting 10×–50× cost jumps. A default “downgrade to base model on budget exhaustion” toggle would preserve continuity and defuse most of the bill-shock anxiety the change created.

Key takeaways

1. Exempt the staple workflow from metering. GitHub keeps code completions unlimited and unbilled so the most frequent interaction never causes bill shock — meter only the expensive, high-variance work.
2. Denominate usage units in real money. Pricing a credit at exactly $0.01 makes usage legible and defensible; abstract units that don’t map to dollars erode trust.
3. Split included usage into fixed + variable layers. A Base (fixed) plus Flex (adjustable) allowance lets you absorb upstream cost swings without repricing the headline seat.
4. Pool usage at the buying entity, not the seat. Org-wide credit pools let power users borrow from light users and make per-seat pricing feel generous without raising the rate.
5. Give finance multi-level budget caps. User, cost-center, and enterprise budgets are what make usage-based overage acceptable to buyers wary of runaway bills.

UBP implications

1. Hybrid seat + usage is the default for AI dev tools. Copilot, like Cursor, pairs a predictable seat with a metered pool — confirming that buyers want baseline predictability plus pay-for-power upside.
2. Carve-outs shape perception more than headline price. By exempting completions, Copilot lets most users perceive a flat subscription even though a meter exists underneath — a powerful packaging move for any usage model.
3. Transparent unit economics build durability. A penny-per-credit pass-through ties the vendor’s price to model cost so directly that customers accept usage billing as fair, easing the path for outcome- and consumption-based models across the category.

Sources

• GitHub Copilot plans & pricing (accessed 2026-06-02)
• About individual GitHub Copilot plans (docs) (accessed 2026-06-02)
• Billing for GitHub Copilot in organizations and enterprises (docs) (accessed 2026-06-02)
• Usage-based billing for individuals (docs) (accessed 2026-06-02)
• Usage-based billing for organizations and enterprises (docs) (accessed 2026-06-02)
• Requests in GitHub Copilot — premium requests (docs) (accessed 2026-06-02)
• Budgets for usage-based billing (docs) (accessed 2026-06-02)
• Copilot for Business (accessed 2026-06-02)
• Browse the full pricing blueprint corpus for peer comparisons.

Bottom line

GitHub Copilot in 2026 is a textbook hybrid: a per-seat subscription wrapped around a dollar-denominated AI Credits pool, with the highest-frequency workflow — code completions — deliberately kept unlimited and unmetered. The penny-per-credit pass-through, Base-plus-Flex allowances, and four-level budget controls show how to make usage billing feel fair and governable at developer, team, and enterprise scale. See how it compares across the pricing blueprint corpus.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

GitHub AI Credits replace premium requests

Jun 2026

On June 1, 2026, Copilot moved usage billing from premium requests to GitHub AI Credits (1 credit = $0.01 USD). Individual plans got Base + Flex credit allowances (Pro 1,500, Pro+ 7,000, Max 20,000 total); the new Max tier launched at $100/mo (upgrade-only); Business and Enterprise got pooled allowances (1,900 and 3,900 per user) with $0.01/credit overage. Existing annual Pro/Pro+ stayed on legacy premium-request billing.

Usage-based-billing migration announced

Apr 2026

GitHub announced (April 27, 2026) that premium request units would be replaced by GitHub AI Credits on June 1, 2026 and that annual Pro/Pro+ plans would be retired at term end. The announcement drew 767 points / 553 comments on Hacker News and press reports of projected 10×–50× agentic cost increases for power users.

New individual sign-ups paused

Apr 2026

From April 20, 2026, new sign-ups for Copilot Pro, Pro+, Max, and student plans were temporarily paused to protect experience quality; existing customers could still upgrade between tiers. The change drew 544 points / 230 comments on Hacker News.

Premium-request billing enforced

Jun 2025

Enforcement of premium-request limits and paid overage began June 18, 2025 on GitHub.com (August 1, 2025 on GHE.com), after the cap-enforcement date slipped from June 4. Paid plans kept their allowances (300 Pro, 1,500 Pro+) with $0.04 per additional request.

Pro+ tier ($39) and premium requests introduced

Apr 2025

Copilot Pro+ launched at $39/mo ($390/yr) and premium-request metering appeared: Free 50, Pro 300, Pro+ 1,500 monthly premium requests, with additional requests at $0.04 each and per-model multipliers. Archived pricing pages from May 2025 confirm the new tier and meter.

Free tier launches; Individual renamed Pro

Dec 2024

GitHub added a $0 Free plan (2,000 completions + 50 chat/agent requests per month) and renamed Copilot Individual to Copilot Pro (still $10/mo). The archived pricing page shows the Free/Pro/Business/Enterprise grid by January 2025.

Copilot Business ($19) and Enterprise ($39) ship

Feb 2024

By February 2024 the org tiers were live: Copilot Business at $19/user/mo and Copilot Enterprise at $39/user/mo (GitHub Enterprise Cloud only), alongside the $10 Individual plan — a clean three-step seat ladder with no usage metering.

General availability at $10/mo for individuals

Dec 2021

Copilot reached general availability for individuals in late 2021 at $10/month or $100/year, with free access for verified students, teachers, and popular open-source maintainers. This single individual tier persisted unchanged through 2024.

Copilot launches in technical preview

Jun 2021

GitHub Copilot launched as a free technical preview on June 29, 2021. The launch (2,905 points / 1,272 comments on Hacker News) was immediately followed by a 'code laundering' controversy over training on public GitHub code.