tech calculator

CI/CD Minutes Cost Calculator

Price GitHub Actions minutes with per-job rounding, OS multipliers, and the included allowance on your plan.

Billed Minutes = Runs × Jobs × the minutes per job rounded up to a whole minute; included minutes drain at an OS multiplier (Linux 1×, Windows 2×, macOS 10×) and overage bills at the runner rate. Example: 200 runs of 4 jobs at 6.5 minutes bills 5,600 minutes and costs $15.60/month on Linux against a 3,000-minute allowance — but $328.60 on macOS.

Formula Verified by Dr. James Callahan, PhD

Quick actions: Save inputs, view saved runs, print, share, or export results

Saved Runs

Inputs

Runner Type

Rate per Minute$

Included-Minute Multiplier×

Plan

Included Minutes per Month

Pipeline Runs per Month

Jobs per Run

Minutes per Jobmin

Results

Monthly Cost

$15.60

Annual Cost$187.20

Cost per Pipeline Run$0.08

Actual Compute Time5,200 min

Billed Minutes (each job rounded up)5,600 min

Lost to Per-Job Rounding400 min

Included Allowance Consumed5,600 min (187% of 3,000)

Chargeable Minutes2600 min

Cost With No Included Allowance$33.60

Tip

Re-run it whenever you add a job, change runner type, or your merge volume climbs — CI bills grow quietly and are almost never budgeted.

Formula

CI/CD minutes cost formula

Billed Minutes = Runs × Jobs per Run × ⌈Minutes per Job⌉
Allowance Consumed = Billed Minutes × OS Multiplier
Chargeable Minutes = max(0, Allowance Consumed − Included Minutes) ÷ OS Multiplier
Monthly Cost = Chargeable Minutes × Rate per Minute

OS multipliers: Linux 1×, Windows 2×, macOS 10×

Two things separate a real CI bill from a naive minutes-times-rate estimate. Each job is billed rounded up to the whole minute, so a workflow made of many short jobs is charged for substantially more than it runs. And included minutes drain at an OS multiplier while overage is billed at the runner's own per-minute rate — which is why a macOS pipeline costs so much more than the rate card alone suggests: it exhausts the allowance ten times faster and then bills at roughly ten times the Linux rate.

Step by step

How the calculation works

1Enter runs per month, the number of jobs in a workflow, and the average minutes a single job takes. Jobs matter separately from total time because the rounding is applied per job, not per workflow — this is the input most estimates collapse and get wrong.
2Pick the runner type to load its published rate and multiplier, or choose Custom and enter your own for a larger runner or a different provider. As of 2026, standard GitHub-hosted rates are $0.005/min for Linux ARM, $0.006 for Linux, $0.010 for Windows, and $0.062 for macOS, with larger Linux runners from $0.012/min at 4 vCPU up to $0.162/min at 64 vCPU.
3Set the plan to load its included minutes: 2,000 a month on Free, 3,000 on Pro and Team, and 50,000 on Enterprise. Included minutes apply to GitHub-hosted standard runners only, and unused minutes do not roll over.
4Watch the rounding line. A pipeline of eight 30-second jobs runs for four minutes and bills for eight — consolidating short jobs into fewer longer ones is often a larger saving than any caching change, because it attacks the rounding rather than the runtime.
5The other structural lever is the runner. Moving a job that does not need macOS or Windows onto Linux typically cuts its cost by an order of magnitude. If your pipeline is heavy enough that self-hosting looks tempting, the [VPS & server cost calculator](/calculators/vps-server-cost-calculator) prices the always-on machine to set against this bill, and the [serverless cost calculator](/calculators/serverless-cost-calculator) covers per-invocation compute for the same comparison.

Related guides

Learn more about this calculation

Guide

How to Use Do The Calculation Calculators: A Practical Step-by-Step Guide

Learn the fastest reliable workflow for using Do The Calculation calculators, reading results, checking formulas, and using save, print, share, and export actions correctly.

Guide

Understanding Calculator Formulas: How DTC Turns Inputs into Results

Understand how Do The Calculation formulas are presented, what the explanation blocks mean, and how to verify calculator logic before using a result in a real decision.

Tech Tool

VPS & Server Cost Calculator

Price a self-hosted runner against your hosted CI bill, including backups and support.

Tech Tool

Serverless Cost Calculator

Compare per-invocation compute pricing against per-minute CI billing.

Business Tool

Software TCO Calculator

Fold CI spend into the wider multi-year cost of the tooling around it.

Tech Tool

Cloud Egress Cost Calculator

Price the artifact and container image transfer your pipeline generates alongside compute.

tech calculator

CI/CD Minutes Cost Calculator

Price GitHub Actions minutes with per-job rounding, OS multipliers, and the included allowance on your plan.

Formula Verified by Dr. James Callahan, PhD

Quick actions: Save inputs, view saved runs, print, share, or export results

Saved Runs

Inputs

Runner Type

Rate per Minute$

Included-Minute Multiplier×

Plan

Included Minutes per Month

Pipeline Runs per Month

Jobs per Run

Minutes per Jobmin

Results

Monthly Cost

$15.60

Annual Cost$187.20

Cost per Pipeline Run$0.08

Actual Compute Time5,200 min

Billed Minutes (each job rounded up)5,600 min

Lost to Per-Job Rounding400 min

Included Allowance Consumed5,600 min (187% of 3,000)

Chargeable Minutes2600 min

Cost With No Included Allowance$33.60

Tip

Re-run it whenever you add a job, change runner type, or your merge volume climbs — CI bills grow quietly and are almost never budgeted.

Summary

Key output metrics

Monthly Cost

$15.60

2600 minutes beyond the included allowance

Annual Cost

$187.20

$0.08 per pipeline run

Allowance Consumed

187%

5,600 of 3,000 included minutes

D3 visualization

Monthly Cost by Runner Type

The identical pipeline priced on each GitHub-hosted runner. macOS is the outlier twice over — ten times the allowance drain and roughly ten times the per-minute rate.

Guide

How to understand and use this calculator

Quick Answer — How Is CI/CD Minute Billing Calculated?

Three steps, and the middle one is where estimates go wrong. Billed Minutes = Runs × Jobs × ⌈ Minutes per Job ⌉, because every job is rounded up to a whole minute. Those billed minutes then drain the included allowance at an OS multiplier, and whatever exceeds it is charged at the runner's own per-minute rate: Cost = (max(0, Billed × Multiplier - Included))/Multiplier × Rate.

GitHub's 2026 published rates for standard hosted runners, with the multiplier at which each drains your included minutes:

Linux ARM — $0.005/min, 1× multiplier
Linux x86 — $0.006/min, 1×
Windows — $0.010/min, 2×
macOS — $0.062/min, 10×

Larger Linux runners run from $0.012/min at 4 vCPU up to $0.162/min at 64 vCPU, all at a 1× multiplier. Included minutes are 2,000 a month on Free, 3,000 on Pro and Team, and 50,000 on Enterprise, and they do not roll over.

The multiplier is the part people miss. A macOS minute costs about ten times a Linux minute *and* burns ten included minutes, so a team on a 3,000-minute plan gets 300 macOS minutes before the meter starts.

How to Use This Calculator: A Busy Linux Pipeline

Enter runs, jobs per run, and average minutes per job separately — the split matters, because rounding is applied per job rather than per workflow. Take 200 pipeline runs a month, 4 jobs each, averaging 6.5 minutes per job, on Linux x86, with a Team plan's 3,000 included minutes.

Actual compute time = 200 × 4 × 6.5 = 5,200 minutes. Billed minutes = 200 × 4 × ⌈6.5⌉ = 200 × 4 × 7 = 5,600 minutes, so 400 minutes a month disappear into rounding before anything else happens.

At a 1× Linux multiplier, those 5,600 billed minutes consume 5,600 of the allowance — 187% of a 3,000-minute plan. Chargeable minutes = 5,600 − 3,000 = 2,600, at $0.006 = $15.60 a month, or $187.20 a year, which works out to about $0.08 per pipeline run.

Now change one thing. Run the identical pipeline on macOS and it still bills 5,600 minutes — but they drain the allowance at 10×, consuming 56,000 minutes against a 3,000-minute plan, leaving 5,300 real minutes chargeable at $0.062. That is $328.60 a month, over twenty times the Linux figure for exactly the same work.

A Second Example: When Rounding Is the Whole Bill

Short jobs are where per-job rounding stops being a footnote. Take a fan-out workflow: 500 runs a month, 8 parallel jobs each, every job finishing in 30 seconds, on Linux with a Free plan's 2,000 included minutes.

Actual compute time = 500 × 8 × 0.5 = 2,000 minutes — exactly the included allowance, so a naive estimate says this pipeline is free. Billed minutes = 500 × 8 × ⌈0.5⌉ = 4,000 minutes. Every 30-second job is charged as a full minute, so 2,000 minutes, half the bill, is pure rounding.

Chargeable minutes = 4,000 − 2,000 = 2,000, at $0.006 = $12.00 a month. Small in absolute terms, but the pipeline consumes 200% of the allowance while doing 100% of the work, and every one of those dollars buys nothing.

The fix is structural rather than technical: consolidate short jobs. Merging those eight 30-second jobs into two two-minute jobs does the same work, bills 2,000 minutes instead of 4,000, and costs nothing at all. No caching change, no faster runner, no reduction in what the pipeline actually does.

Where CI Spend Actually Goes, and How to Cut It

In rough order of impact, four levers matter. Runner choice is the largest and the easiest: any job that does not genuinely need macOS or Windows belongs on Linux, and moving it typically cuts that job's cost by an order of magnitude. macOS is unavoidable for iOS builds and signing, but it is frequently used for jobs that would run identically on Linux simply because the workflow was copied from one that needed it.

Job granularity is second, and counterintuitive. Splitting a workflow into many small parallel jobs feels like an optimisation — it does reduce wall-clock time — but each job pays the rounding penalty separately. Parallelism that halves your wait while doubling your bill is a trade worth making deliberately, not by accident.

Trigger discipline is third. Pipelines that run on every push to every branch, rather than on pull requests and merges, routinely multiply monthly minutes several times over for no additional signal. Path filters and concurrency groups that cancel superseded runs are usually the highest-value configuration change available.

Caching is fourth, and it is where most teams start — not wrong, just less leveraged than the three above. Dependency caching genuinely reduces minutes per job; it just cannot fix a pipeline running on the wrong runner, sliced into 30-second fragments, on every push. If your volume is high enough that self-hosting starts to look attractive, price the alternative honestly with the VPS server cost calculator — an always-on machine bills for its idle hours too, so the crossover point is further out than it looks.

Limitations

This calculator performs exact arithmetic on the runs, jobs, minutes, rate, multiplier, and allowance you enter. The default rates and multipliers follow GitHub's published 2026 GitHub-hosted runner pricing, but CI pricing has changed repeatedly and providers restructure tiers — check the current pricing page before budgeting, and use the custom runner option for anything not listed.

It models compute minutes only, which is not the whole CI bill. Artifact and log storage, container registry storage and transfer, cache storage beyond the free allowance, and data egress from your pipeline are billed separately and can rival compute on artifact-heavy workflows. Self-hosted runners consume no GitHub-hosted minutes but cost real money in servers, maintenance, and the security work that comes with running your own build infrastructure.

It also assumes an average job length, where real pipelines have a distribution. A workflow whose jobs range from 20 seconds to 20 minutes will not behave like one where every job takes the average, particularly for rounding — the short tail is charged proportionally far more. If your jobs vary widely, run the calculator separately for each cluster of similar jobs and add the results rather than averaging across all of them.

Related Calculators

If your CI volume is high enough to consider self-hosting, the VPS Server Cost Calculator prices the always-on machine to set against this bill, including the backup and support lines that are easy to forget. The Serverless Cost Calculator covers per-invocation compute for the same kind of comparison. To fold CI spend into the wider multi-year cost of your development tooling, the Software TCO Calculator models subscriptions, escalation, and internal admin time together, and the Cloud Egress Cost Calculator prices the artifact and image transfer your pipeline generates alongside compute.

Related concepts

Per-Job Minute Rounding

Each job is billed rounded up to a whole minute, independently. A workflow of many short jobs is charged for substantially more time than it actually runs.

OS Multiplier

How fast a runner drains your included allowance — Linux 1×, Windows 2×, macOS 10×. Separate from, and multiplied on top of, the per-minute overage rate.

Included Minutes

The monthly allowance bundled with a plan for GitHub-hosted standard runners: 2,000 Free, 3,000 Pro and Team, 50,000 Enterprise. They do not roll over.

Frequently Asked Questions

How much does GitHub Actions cost per minute?+

As of 2026, standard GitHub-hosted runners are $0.005/min for Linux ARM, $0.006/min for Linux x86, $0.010/min for Windows, and $0.062/min for macOS. Larger Linux runners range from $0.012/min at 4 vCPU to $0.162/min at 64 vCPU. Those rates apply only to minutes beyond your plan's included allowance.

How many free GitHub Actions minutes do I get?+

2,000 minutes a month on the Free plan, 3,000 on Pro and Team, and 50,000 on Enterprise, for GitHub-hosted standard runners. They do not roll over, and they drain at the OS multiplier — so a 3,000-minute allowance is only 300 macOS minutes.

What are the GitHub Actions OS multipliers?+

Linux drains included minutes at 1×, Windows at 2×, and macOS at 10×. The multiplier controls how fast your allowance is consumed; overage beyond it is billed at the runner's own per-minute rate. macOS is expensive twice over — it burns the allowance ten times faster and then bills at roughly ten times the Linux rate.

Does GitHub round CI minutes up?+

Yes, and per job rather than per workflow. A job that runs for 30 seconds is billed as a full minute, so a workflow of eight 30-second jobs runs for four minutes and bills for eight. On short-job pipelines this rounding can be half the entire bill.

How do I reduce my GitHub Actions bill?+

In order of impact: move any job that does not need macOS or Windows onto Linux; consolidate very short jobs so you pay the rounding penalty fewer times; restrict triggers so pipelines run on pull requests and merges rather than every push, with concurrency groups cancelling superseded runs; and cache dependencies. Most teams start with caching, which is the least leveraged of the four.

Is parallelising my CI jobs cheaper?+

No — it is usually more expensive, though faster. Splitting work across more jobs reduces wall-clock time but pays the per-job minute rounding separately on each one. That can be a good trade for developer productivity, but it should be a deliberate choice rather than an assumed optimisation.

Are self-hosted runners free?+

They consume no GitHub-hosted minutes. GitHub proposed a $0.002/min self-hosted platform fee for March 2026 and then postponed it indefinitely, so as of this writing self-hosted minutes remain free of that charge. They are not free in reality: servers, maintenance, and the security work of running your own build infrastructure all cost money, which is why the crossover point is further out than raw minute pricing suggests.

Does this include artifact and cache storage costs?+

No. This models compute minutes only. Artifact and log storage, container registry storage and transfer, cache beyond the free allowance, and pipeline egress are billed separately and can rival compute cost on artifact-heavy workflows. Add those from your provider's storage pricing.

Formula

CI/CD minutes cost formula

Billed Minutes = Runs × Jobs per Run × ⌈Minutes per Job⌉
Allowance Consumed = Billed Minutes × OS Multiplier
Chargeable Minutes = max(0, Allowance Consumed − Included Minutes) ÷ OS Multiplier
Monthly Cost = Chargeable Minutes × Rate per Minute

OS multipliers: Linux 1×, Windows 2×, macOS 10×

Step by step

How the calculation works

1Enter runs per month, the number of jobs in a workflow, and the average minutes a single job takes. Jobs matter separately from total time because the rounding is applied per job, not per workflow — this is the input most estimates collapse and get wrong.
2Pick the runner type to load its published rate and multiplier, or choose Custom and enter your own for a larger runner or a different provider. As of 2026, standard GitHub-hosted rates are $0.005/min for Linux ARM, $0.006 for Linux, $0.010 for Windows, and $0.062 for macOS, with larger Linux runners from $0.012/min at 4 vCPU up to $0.162/min at 64 vCPU.
3Set the plan to load its included minutes: 2,000 a month on Free, 3,000 on Pro and Team, and 50,000 on Enterprise. Included minutes apply to GitHub-hosted standard runners only, and unused minutes do not roll over.
4Watch the rounding line. A pipeline of eight 30-second jobs runs for four minutes and bills for eight — consolidating short jobs into fewer longer ones is often a larger saving than any caching change, because it attacks the rounding rather than the runtime.
5The other structural lever is the runner. Moving a job that does not need macOS or Windows onto Linux typically cuts its cost by an order of magnitude. If your pipeline is heavy enough that self-hosting looks tempting, the [VPS & server cost calculator](/calculators/vps-server-cost-calculator) prices the always-on machine to set against this bill, and the [serverless cost calculator](/calculators/serverless-cost-calculator) covers per-invocation compute for the same comparison.

Related guides

Learn more about this calculation

Guide