Summary: GitHub Copilot no longer counts "premium requests" for most people. Every individual plan now comes with a monthly pool of GitHub AI Credits (1 credit = $0.01), and what a chat or agent task costs depends on the model and the number of tokens it uses. This guide breaks down what Free, Pro ($10), Pro+ ($39) and Max ($100) actually include, how the "flex allotment" works, and how to make sure you never pay more than the subscription price.
What changed: from request counts to AI credits
If you read older Copilot guides, you will see talk of a fixed number of premium requests per month and model "multipliers". GitHub's docs now label that system as legacy: it only applies to Copilot Pro and Pro+ subscribers on an existing annual plan who stayed on request-based billing after June 1, 2026. Everyone else is on usage-based billing measured in GitHub AI Credits.
The new rule is simple to state and a little harder to predict. Each interaction uses input tokens (what is sent to the model), output tokens (what the model writes back) and cached tokens (context the model reuses). Each model has its own per-token price, the total is converted into credits, and one credit equals one US cent. A quick question to a lightweight model can cost a fraction of a credit; a long cloud agent session on a frontier model across many files costs far more.
The good news for everyday coding: code completions and next edit suggestions are not billed in credits and stay unlimited on all paid plans. Credits are spent by Copilot Chat, Copilot CLI, the Copilot cloud agent, Copilot Spaces, Spark and third-party coding agents.
The plans side by side
These figures come from GitHub's Copilot plans page and the "Plans for GitHub Copilot" and individual billing articles in GitHub Docs. "Base credits" match your subscription price; the "flex allotment" is extra included usage on top of that.
| Plan | Price | Monthly AI credits (base + flex) | Code completions | Models and key extras |
|---|---|---|---|---|
| Copilot Free | $0, no credit card | A limited allowance (GitHub gives no number); you cannot buy extra credits | 2,000 per month | Auto model selection only; agent mode, MCP and Copilot CLI included; no Copilot in the cloud |
| Copilot Pro | $10/month (free for verified teachers and some open source maintainers) | 1,000 + 500 = 1,500 (about $15 of model usage) | Unlimited | Choose from a selection of models; cloud agent and code review; third-party agents (Claude Code, Codex) in public preview |
| Copilot Pro+ | $39/month | 3,900 + 3,100 = 7,000 (about $70) | Unlimited | Adds premium models such as Opus; audit logs |
| Copilot Max | $100/month | 10,000 + 10,000 = 20,000 (about $200) | Unlimited | Everything in Pro+ plus priority access to new models and features |
There is also Copilot Student, which is free for verified students, includes unlimited code completions and an allowance of credits, uses auto model selection, and leaves out third-party agents. For teams, Copilot Business is $19 per seat per month with 1,900 credits per user and Copilot Enterprise is $39 per seat with 3,900 credits per user; those credits go into a shared pool and usage beyond it is charged at $0.01 per credit.
How the flex allotment works (and why it matters)
Your base credits are always spent first. Once they are gone, the flex allotment kicks in automatically at the same rates in your IDE, on GitHub.com and in Copilot CLI. You do not have to switch anything on.
The catch is in the wording. GitHub describes base credits as fixed ("they never change"), while the flex allotment is "a variable part of your included usage" designed to adapt to model prices and efficiency, and the plans page footnote says flex allotments will vary over time. So the honest way to compare plans is: Pro guarantees 1,000 credits, Pro+ 3,900 and Max 10,000; the extra 500, 3,100 and 10,000 are current figures, not promises.
Two more timing details are easy to miss:
- Unused credits do not roll over. Your allowance resets at 00:00:00 UTC on the first day of every calendar month, no matter which day you subscribed. In UTC+9 time zones such as Korea and Japan, that is 9:00 a.m. on the 1st.
- Your billing date is separate. If you are not already paying GitHub for something, your billing cycle usually starts on the day you sign up, but the credit reset stays on the 1st.
What a credit buys: reading the model price table
GitHub publishes per-token prices for every model in the "Models and pricing for GitHub Copilot" reference, quoted per 1 million tokens. A few examples from that page show how wide the range is:
- GPT-6 Luna (lightweight): $0.10 input, $0.50 output.
- Claude Sonnet 5 (versatile): $2.00 input, $10.00 output.
- Claude Opus 5 (powerful): $5.00 input, $25.00 output.
- GPT-6 Astra (powerful): $10.00 input, $50.00 output, doubling to $20 and $75 above 272K input tokens.
To make that concrete, here is some simple arithmetic using those published prices (an illustration, not a measured session). Suppose one agent turn sends 50,000 input tokens and gets 2,000 output tokens back. On Claude Sonnet 5 that is $0.10 + $0.02 = $0.12, or 12 credits. The same turn on GPT-6 Luna is $0.005 + $0.001, well under one credit. On GPT-6 Astra it is $0.50 + $0.10 = 60 credits. In practice, cached input is priced much lower than fresh input, so a long session that keeps its cache warm costs less than this naive sum, while one that keeps rebuilding context costs more.
The takeaway: Pro's 1,500 credits can last a whole month if you mostly use completions and light chat, but a few long agent sessions on a frontier model can use them up in days. Model choice matters more than plan choice.
There are two discounts worth knowing: paid plans get 10% off model costs when using auto model selection in Copilot Chat or the cloud agent, and through November 30, 2026, paid users get 30% off model costs when using Copilot Fusion, which picks models per task.
Step by step: cap your spending at the plan price
When your included credits run out on an individual plan, you have three options: upgrade (you pay only the price difference, and your earlier usage counts against the bigger allowance), keep your plan and pay for additional usage, or wait for the monthly reset. If you want the third option guaranteed, set a $0 budget. GitHub's docs recommend this exact setup to avoid charges beyond your plan:
- Open your billing overview at github.com/settings/billing (sign in first).
- Click Budgets and alerts, then New budget.
- Under "Budget Type", choose Bundled AI credits budget. This covers every SKU that consumes AI credits, not only Copilot Chat.
- Set the scope, then set the amount to $0.
- Tick Stop usage when budget limit is reached. Without it, GitHub only emails you when you go over; usage is not stopped.
- Optionally tick Receive budget threshold alerts to get emails at 75%, 90% and 100%, then click Create budget.
On the same page, under "Included usage alerts", you can also opt in to emails when your included plan usage hits 90% and 100%. That is the earlier warning most people actually want.
Cap a single agent task in Copilot CLI
A budget protects the month; a session limit protects one task. In an interactive Copilot CLI session, run /limits set max-ai-credits 200 (any number of at least 30) and Copilot stops cleanly when the session reaches it, letting you raise the limit and continue. For scripts, pass --max-ai-credits with copilot -p "YOUR PROMPT". GitHub notes this feature is in public preview and is a soft limit: a response already in progress finishes, so usage can go slightly over.
How to choose a plan
Stay on Free if
You mainly want tab completions and the occasional chat question. 2,000 completions a month is enough for light or hobby coding, and you cannot be charged because Free does not allow buying extra credits.
Pick Pro if
You code most days and want unlimited completions plus model choice, the cloud agent and code review. Budget around light-to-moderate agent use on mid-priced models, and keep heavy reasoning models for planning.
Pick Pro+ or Max if
You need Opus-class or other premium models, or you run long agent sessions daily. Upgrading mid-month only charges the difference, so it is reasonable to start on Pro and move up when the 90% alert starts arriving every month.
Editor's note
Who this is for: individual developers deciding between Free, Pro, Pro+ and Max, and anyone who saw "AI credits" in their billing page and wants to know what changed.
Who it is not for: people whose Copilot comes from an employer. Business and Enterprise seats use a pooled allowance set by admins; if you run out you will see a banner offering to request more budget from your administrator. Also note that being assigned an organization seat automatically cancels a personal Pro, Pro+ or Max plan, with a prorated refund.
Gotchas I found while checking the docs: the flex part of your allowance can change; credits reset on the 1st in UTC, not on your billing date; switching models mid-session breaks the prompt cache, so the next request is billed as fresh input; and if you bought Pro through the iOS or Android app, you cannot buy extra credits in the app store.
Common confusion and the fix: if your usage numbers or model prices look wrong in your editor, GitHub says older clients may show outdated billing terms. Update to at least VS Code 1.120, JetBrains plugin 1.9.1 or Copilot CLI 1.0.48. If you hit your limit and Copilot will not continue, either raise your budget above $0, upgrade, or wait for the reset.
FAQ
Do code completions use GitHub AI Credits?
No. Code completions and next edit suggestions are not billed in credits and are unlimited on paid plans. The Free plan caps completions at 2,000 per month instead.
Can I get charged more than $10 a month on Copilot Pro?
Only if you allow additional usage. Create a Bundled AI credits budget of $0 with "Stop usage when budget limit is reached" enabled, and Copilot stops instead of billing you once your included credits are used.
I am on an annual plan. Do AI credits apply to me?
Maybe not yet. GitHub's legacy requests article applies to Pro and Pro+ subscribers on an existing annual plan who remained on premium request billing after June 1, 2026. Check your plan under Settings, Billing and licensing, Licensing.
Related guides on this blog
- Claude Code Errors Fixed: Login, PATH, 429 and Usage Limits
- Codex CLI Not Working? Fixes for Login, Sandbox and Limits
- Cline Tutorial: Plan, Act and Roll Back Safely in VS Code
Treat the base credits as your real allowance, set a $0 budget on day one, and let your monthly usage alerts tell you when it is time to upgrade.
Images: own screenshots of GitHub's official Copilot plans page and GitHub Docs (github.com, docs.github.com); see them at the end of this post.
Last verified: 2026-11-03
Sources:
GitHub Copilot plans and pricing
GitHub Docs: Plans for GitHub Copilot
GitHub Docs: Usage-based billing for individuals
GitHub Docs: Models and pricing for GitHub Copilot
GitHub Docs: Setting up budgets to control spending
GitHub Docs: Setting an AI credit session limit in Copilot CLI
GitHub Docs: Optimizing your AI usage
GitHub Docs: Requests in GitHub Copilot (legacy)
댓글
댓글 쓰기