GitHub Copilot Pricing: Plans, AI Credits & What Changed
This GitHub Copilot pricing 2026 guide covers the platform’s biggest overhaul yet. The old “premium requests” system is gone, replaced by usage-based billing measured in GitHub AI Credits. Plan prices stayed the same — but what your subscription actually buys changed completely.
This guide covers every Copilot plan in 2026, explains how AI Credits work, breaks down the June 2026 billing change, and helps you figure out what you will really pay each month.
GitHub Copilot Pricing 2026: Plans at a Glance
| Plan | Price | Included AI Credits | Best for |
|---|---|---|---|
| Copilot Free | $0 | Limited | Trying Copilot, light use |
| Copilot Pro | $10/month | $10 in credits (1,000 credits) | Individual developers |
| Copilot Pro+ | $39/month | $39 in credits (3,900 credits) | Heavy users, premium models |
| Copilot Max | $100/month | Higher allowance | All-day agentic coding |
| Copilot Business | $19/user/month | $19 per user, pooled | Teams |
| Copilot Enterprise | $39/user/month | $39 per user, pooled | Large organizations |
One credit equals roughly $0.01, so $10 in credits is 1,000 credits. Annual plans have been retired — existing annual subscribers are grandfathered until their term expires, then move to monthly or Free.
For the official details, GitHub documents the full model in its billing docs: About billing for GitHub Copilot, and the plan lineup lives at github.com/features/copilot.
What Are GitHub AI Credits?
GitHub AI Credits are the new metering currency. Instead of counting “requests” as identical units, GitHub now measures actual token consumption — input tokens, output tokens, and cached tokens — at each model’s published per-token API rate, then converts that dollar cost into credits.
Why the change? Under the old system, a quick chat reply and a 20-minute autonomous agent run each cost one “premium request,” even though the agent run could consume a hundred times the compute. That mismatch became unsustainable as agentic coding exploded in late 2025 and 2026. Token-based billing aligns what you pay with what you actually consume.
The practical consequences:
- Model choice now affects your bill directly. A long agent session on a frontier model burns credits far faster than quick chats on a small model.
- Cached tokens are cheaper. Repeated context (like your repo’s structure) costs less than fresh input.
- You can run out mid-month. Once your included credits are gone, you either set an overflow budget for automatic top-ups or wait until the next cycle.
The Big June 2026 Change: Premium Requests → AI Credits
On June 1, 2026, the switch flipped. Here is the before and after:
| Old model (until May 31, 2026) | New model (from June 1, 2026) | |
|---|---|---|
| Currency | Premium requests | GitHub AI Credits |
| Unit | 1 request = 1 unit, any size | Tokens × per-model rate |
| Pro allowance | 300 premium requests/month | $10 in AI credits |
| Pro+ allowance | 1,500 premium requests/month | $39 in AI credits |
| Overage | $0.04 per request | Per-token billing |
| Seat prices | Unchanged | Unchanged |
To ease the transition, existing Business and Enterprise customers received promotional bonus credits through August 2026 (roughly 1.5–1.8× their standard allotment).
Individual Pro and Pro+ plans also include a flex allotment on top of the base credits — GitHub says this flex amount may shift as underlying model prices change, so treat the headline numbers as the guaranteed floor.
What Is Still Unlimited (and Free)
Here is the good news, and it is significant: inline code completions and Next Edit Suggestions remain unlimited on all paid plans and never consume AI Credits.
If your Copilot usage is mostly what Copilot was originally famous for — ghost-text completions as you type, accepting suggestions line by line — the June 2026 change barely touches you. Your $10 Pro subscription still covers unlimited completions exactly as before.
The metered model applies to the newer, heavier features:
- Copilot Chat
- Agent mode / agentic coding sessions
- Copilot CLI
- Copilot code review
- Copilot cloud agents
- Copilot Spaces and GitHub Spark
- Third-party Copilot extensions
Think of it as two products sharing one subscription: the classic autocomplete (unlimited) and the agentic AI pair-programmer (metered).
Copilot Free: What $0 Gets You in 2026
The Free plan is a genuine trial tier, not a usable daily driver:
- Limited AI Credits each month
- Around 2,000 code completions per month
- Access to chat and agent features within the credit cap
It is enough to evaluate whether Copilot fits your workflow. The moment you rely on agent mode for real work, you will want Pro.
Copilot Pro vs Pro+ vs Max: Which Individual Plan?
Pro ($10/month) is the sweet spot for most developers. You get unlimited completions plus $10 in monthly AI Credits — enough for regular chat usage and occasional agent sessions. If you find yourself rationing agent runs in the last week of the month, that is your signal to upgrade.
Pro+ ($39/month) quadruples your credit allowance to roughly $39 and unlocks premium models (including Opus-class models) that are excluded from Pro’s base allowance. It also adds features like Spark. Choose Pro+ if you routinely work with the most capable models or run long agent sessions.
Max ($100/month) is for developers who effectively run Copilot as an autonomous engineer — hours of agentic coding per day, cloud agents working in parallel. At this level you also get priority access to new models. For almost everyone else, it is overkill; the jump from Pro+ to Max only makes sense when your credit consumption consistently exceeds ~$39/month.
Copilot Business vs Enterprise: The Team Tiers
Business ($19/user/month) adds what teams need: SSO, admin and policy controls, audit logs, IP indemnity, training exclusion, and a data processing agreement. Credits pool across the organization, so a light user’s unused allowance covers a heavy user’s overage — this pooling is genuinely valuable and often overlooked in cost planning.
Enterprise ($39/user/month) doubles the per-seat credit allowance and adds SAML SSO, fine-tuning options, and a data-residency path. Organizations subject to strict compliance requirements (or the +10% multiplier for FedRAMP/data-residency routing) will land here.
Admins can cap spending at the enterprise, cost-center, or individual user level, which is the control you want configured before rolling Copilot out to 500 developers.
How Much Will You Actually Spend? Three Scenarios
The autocomplete developer. Uses Copilot mainly for inline completions, occasional chat questions. Credit usage: minimal. Cost: $10/month (Pro). The June change is nearly invisible.
The agent-mode enthusiast. Runs several agent sessions per week, uses chat daily, occasionally reaches for premium models. Credit usage: moderate, maybe $15–25/month in tokens. Cost: $10/month Pro + small overflow, or $39/month Pro+ for headroom and premium models.
The team of 20. Mixed usage, some heavy agent users. Pooled Business credits ($19 × 20 = $380 in monthly credits) smooth out the variance. Cost: $380/month, with admin caps preventing surprises.
The honest takeaway: for most individuals, the bill did not go up — it got more transparent. Heavy agent users are the ones who need to watch the meter.
How Copilot Pricing Compares
Copilot’s $10 entry point remains one of the cheapest ways to get daily AI coding assistance, especially with unlimited completions included. But the market has options:
Prefer an AI-first editor over an extension? Check our breakdown of Cursor's free plan and paid tiers.
Microsoft is also pushing autonomous agents beyond coding — our Microsoft Agent 365 overview explains what it is and how its pricing works. Microsoft Agent 365 overview and pricing
- If you are comparing AI assistants broadly — not just for code — our ChatGPT vs Copilot breakdown looks at which tool actually fits your work.
- For pure API-style consumption pricing, see our Google Gemini API pricing guide for 2026 — token-metered billing is the industry’s direction, and Copilot’s credit system follows the same logic.
- And DeepSeek’s API pricing shows how aggressively the low-cost providers are pricing the same underlying tokens.
Tips to Keep Your AI Credit Bill Down
- Default to smaller models for chat and routine agent tasks; reserve frontier models for genuinely hard problems.
- Let completions do the work. Inline suggestions are unlimited — lean on them before opening an agent session.
- Batch your agent runs. One well-specified session beats five exploratory ones.
- Watch the meter. Check credit consumption weekly for the first two months to learn your pattern.
- Use pooled plans for teams. Business/Enterprise pooling absorbs the variance between light and heavy users.
- Set overflow budgets deliberately. Decide in advance whether overage auto-bills or stops — do not discover the default mid-deadline.
Frequently Asked Questions
Is GitHub Copilot free?
There is a Free plan with limited credits and completions. Full use requires a paid plan starting at $10/month.
What happened to premium requests?
They were retired on June 1, 2026, replaced by GitHub AI Credits billed on actual token usage. Plan prices did not change.
Do code completions use AI Credits?
No. Inline completions and Next Edit Suggestions are unlimited on paid plans and never consume credits.
How much is one GitHub AI Credit worth?
Approximately $0.01. So 1,000 credits ≈ $10.
Can I still buy an annual Copilot plan?
No — annual plans were retired. Existing annual subscribers keep their terms until expiry, then move to monthly billing or Free.
What uses the most credits?
Long autonomous agent sessions on frontier models — they combine high token volumes with the most expensive per-token rates. Monitor these specifically.
Final Thoughts
GitHub Copilot in 2026 costs the same on paper — $0 to $100 for individuals, $19 or $39 per seat for teams — but the June shift to token-metered AI Credits changed the strategy. Completions stay unlimited, which protects the core experience, while chat, agents, and the CLI now draw from a visible meter. Learn your consumption pattern in the first month, pick models deliberately, and check the official billing docs before you scale up. For most developers, the new model is not more expensive — it is just more honest.
Copilot CLI: The Terminal Counts Too
One of the biggest practical changes in 2026 is that the Copilot CLI — GitHub’s terminal-based coding agent — draws from the same AI Credit pool as chat and agent mode. Every CLI session meters tokens exactly like an IDE agent run.
This matters because terminal workflows tend to be long-running: you kick off a refactor, the agent explores the repo, runs tests, iterates. A single CLI session can consume more credits than a full day of chat. If you adopt the CLI, treat it as the heaviest item in your Copilot budget and monitor its consumption separately for the first few weeks. The upside is real — repo-wide refactors from the terminal are genuinely powerful — but “fire and forget” is also the fastest way to burn a month of credits in an afternoon.
Copilot Code Review: Metered, Plus Actions Minutes
Automated code review is one of Copilot’s most popular 2026 features — and it has a double cost most people miss. Review sessions consume AI Credits like any agent workload, and they also consume GitHub Actions minutes, because reviews run on Actions runners.
For a solo developer reviewing a few PRs a week, this is negligible. For a team with dozens of daily PRs and review-on-every-push enabled, both meters move meaningfully. If your Actions bill suddenly grows alongside Copilot adoption, the code review feature is the first place to look. Consider scoping automatic reviews to specific branches or file paths rather than every push.
Free Access: Students, Educators & Open-Source Maintainers
GitHub continues its education and open-source programs in 2026:
- Students and educators verified through GitHub Education get Copilot Pro free.
- Maintainers of popular open-source projects can qualify for free access.
If you are eligible and paying for Pro, stop — verify your status and claim the benefit. These programs are the best deal in the Copilot lineup, and a surprising number of eligible developers never apply.
Migrating From the Old Plan: What Existing Subscribers Should Do
If you subscribed before June 2026, here is your checklist:
- Nothing broke. Your seat price is the same; only the metering changed.
- Check your first two invoices. Compare token-metered consumption against your old premium-request usage to spot behavior changes.
- Re-evaluate Pro vs Pro+. Under premium requests, heavy chat users sometimes needed Pro+ for the request allowance. Under token billing, chat on efficient models is cheap — you might drop to Pro, or you might need Pro+ for premium-model access instead. The decision criteria changed.
- Configure admin caps now (Business/Enterprise). Token billing makes per-user variance larger, so guardrails matter more than they did.
- Audit third-party extensions. Copilot extensions consume credits too — an extension you installed in 2025 and forgot about is now a line item.
Copilot and Your IDE: Does the Editor Change the Price?
No — the plan is the plan regardless of where you use it. VS Code, Visual Studio, JetBrains IDEs, Neovim, the CLI, and the web all draw from the same subscription and credit pool. What changes the cost is what you do, not where: an agent session in Neovim meters identically to one in VS Code.
The one editor-adjacent exception is the cloud agent (fire-and-forget tasks that run in GitHub’s cloud and open PRs). These are among the heaviest credit consumers because they run long, autonomous sessions on capable models. Powerful — but budget them like small batch jobs, not like chat.
Frequently Asked Questions (Continued)
How do I check my remaining AI Credits?
In your GitHub billing settings, under the Copilot section. Team admins see organization-level consumption and per-user breakdowns. Check it weekly for the first month to learn your pattern.
Do premium models cost more credits?
Yes, effectively — credits are consumed based on tokens multiplied by each model’s per-token rate, so frontier models drain your allowance faster than smaller ones. This is the core behavior to internalize.
What happens when I run out of credits mid-month?
You can configure an overflow budget that auto-purchases additional usage, or let usage pause until the next billing cycle. Set this intentionally before you need it.
Is Copilot Business worth it over Pro for a small team?
If you need SSO, audit logs, IP indemnity, or policy controls — yes, those are Business-only. On pure cost, pooled credits make Business efficient for teams with mixed usage. A team of five heavy agent users might actually spend less pooled on Business than on five separate Pro+ seats.
Does Copilot train on my code?
GitHub offers training exclusion controls on Business and Enterprise plans. Individual users should review the current privacy settings for their plan tier — the defaults and options have evolved, so verify rather than assume.
The Bottom Line for 2026
Zoom out, and the story of Copilot pricing in 2026 is the story of the whole AI industry: flat-rate abundance giving way to metered reality. GitHub kept seat prices frozen and completions unlimited — the two decisions that matter most to everyday users — while making heavy agentic usage pay its own way.
Developers who mainly accept ghost text will never notice the change. Developers who live in agent mode now have a fuel gauge, which is ultimately fairer than the alternative (higher flat prices for everyone). Watch your first month of credits, choose models with intent, and Copilot remains one of the best-value AI tools a developer can buy.
