GitHub Copilot AI credits calculator: credits per model, task and month
Estimate how many GitHub AI Credits a Copilot workload uses: pick a model, enter the tokens one model call sends and gets back, and how many calls and tasks you run in a month. You get the credits, the dollar value and the share of your plan's monthly allowance, plus the same workload priced on every model.
Rates are GitHub's per-token prices for each model (default tier, per 1 million tokens), read from GitHub's docs on November 14, 2026. Everything is calculated in your browser. Nothing you enter is sent or stored.
Start from an example (our assumed token counts, not measurements):
Everything sent to the model: your prompt, open files, instructions, tool definitions and, in a conversation, the history so far.
Agent sessions resend most of the same context on every call, so a large share can come from the cache. A new chat question usually has none.
1 for a chat question. An agent session makes one call per step, often dozens.
Cache writes and code review
Many models list a separate rate for storing context in the cache. GitHub does not say how many tokens an interaction writes, so this stays 0 unless you enter it; tokens entered here are priced at the cache-write rate instead of the input rate.
GitHub picks the review model and does not disclose it, so reviews use GitHub's published average per review. Reviews also use GitHub Actions minutes, which are not included here.
Loading the rates...
The same workload on every model
Cheapest first. The share is of one month's allowance on your plan. Your selected model is highlighted.
| Model | GitHub's category | Credits per task | Credits per month | Share of allowance |
|---|
How the estimate works
- Formula: for one call, (uncached input × input rate + cached input × cached rate + cache writes × cache-write rate + output × output rate) ÷ 1,000,000 gives dollars; 1 AI credit is $0.01. A task is that times the number of calls, and the month adds the tasks and the code reviews.
- Long context: some models list higher rates above an input-token threshold (for example 272K for several GPT models, 100K for Claude Haiku 5.5). When one call's input, cached tokens included, is above the threshold, we price the whole call at the long-context rates. GitHub's table states the threshold in input tokens but not whether cached tokens count, so calls near the threshold are approximate.
- Promotional prices are used until the date GitHub gives. After it, a model switches to its announced regular price; if GitHub has not published one, the model is left out instead of showing an old price.
- Discounts: GitHub takes 10% off model costs on paid plans when you use auto model selection in Copilot Chat or Copilot cloud agent. Copilot Fusion takes 30% off for a limited time. Only one discount is applied at a time.
- Not included: code completions and next edit suggestions (not billed in AI credits on paid plans), GitHub Actions minutes used by code review, and any rounding GitHub applies. Your usage page on GitHub is the record of what you actually used.
- Copilot Free and Copilot Student also include AI credits, but GitHub does not publish the amount, so they are not in the plan list.
The arithmetic, the plan allowances and what happens when credits run out are explained step by step in GitHub Copilot AI credits explained. GitHub's own pages: Models and pricing, billing for individuals, billing for organizations.
FAQ
How many AI credits does each GitHub Copilot plan include?
Per GitHub's docs (checked November 14, 2026): Copilot Pro ($10 a month) 1,500 credits, Pro+ ($39) 7,000 and Max ($100) 20,000, each made of base credits plus a flex allotment that GitHub says may change. Copilot Business includes 1,900 and Copilot Enterprise 3,900 credits per user a month, pooled across the organization. Free and Student include an allowance without a published number.
How much is one AI credit worth?
$0.01. A $10 budget for additional usage covers 1,000 AI credits.
Do code completions use AI credits?
No. Code completions and next edit suggestions are not billed in AI credits and stay unlimited on paid plans. Chat, Copilot CLI, Copilot cloud agent, Spaces, Spark and third-party coding agents are.
Why does the cached share change the result so much?
Cached input is priced far below normal input, often at a tenth of it. Agent sessions resend much of the same context on every call, so with a high cache share most of the input is billed at the cached rate. Switching models or settings mid-session, or a cache that has expired, sends the whole context at the full rate again.
What is the long-context rate?
Some models list higher rates when the input of a call is above a threshold, for example 272K tokens for several GPT models and 100K for Claude Haiku 5.5. The calculator prices the whole call at those rates when the input you enter is above the threshold.
How accurate is the estimate?
It applies GitHub's published per-token rates to the token counts you enter, so it is as good as those counts. It leaves out cache writes unless you enter them, GitHub Actions minutes for code review and any rounding on GitHub's side. Your usage page on GitHub is the record of what you actually used.
Is anything I enter sent anywhere?
No. The page loads the rate table from this site and does the arithmetic in your browser. The numbers you enter stay in the page and in the link after # if you copy it.