GitHub larger runners pricing (2026), and a flat monthly price for 4, 8 and 16 vCPU jobs

GitHub's Linux x64 larger runners cost $0.012 a minute at 4 cores, $0.022 at 8 and $0.042 at 16, up to $0.252 at 96 cores, and GitHub "rounds the minutes and partial minutes each job uses up to the nearest whole minute" (GitHub Docs, read on November 9, 2026). They're "only available for organizations and enterprises using the GitHub Team or GitHub Enterprise Cloud plans", they draw on the plan's included compute, and "Larger runners are not free for public repositories."

Skipbox publishes this comparison and sells one of the two runners in it. Every GitHub figure comes from GitHub's own pages, read on November 9, 2026 and linked under Sources.

The price list

Per minute, in US dollars:

The pricing page also lists a Linux 2-core arm64 larger runner ($0.005) and Windows arm64 larger runners from 2 to 64 cores. For scale, the standard ubuntu-latest runner on a private repository is 2-core at $0.006 a minute, so a 16-core minute costs 7 times a standard one.

Who can use them, and how they bill

• GitHub Team or Enterprise Cloud only. An organization owner adds a larger runner to the organization and picks the repositories that may use it. Personal accounts and organizations on GitHub Free have no larger runners.
• A card and a budget. The larger runners reference says: "You must have a valid credit card on file in order to use larger runners", and "Your GitHub Actions spending limit must be set to a value greater than zero."
• Per minute, rounded up per job. A job that runs 3 minutes 10 seconds on a 16-core runner bills 4 minutes, $0.168.
• Included compute first. On Team and Enterprise Cloud, a larger runner's cost "can draw from the included compute usage": $18 a month on Team, $300 on Enterprise Cloud. GitHub counts it in dollars, "not a fixed number of minutes", so $18 buys 1,500 minutes on a 4-core runner and about 428 on a 16-core one.
• Public repositories pay. Standard runners are free there; larger runners aren't.
• Billing SKUs. Each size has its own SKU (linux_8_core and so on) in the usage report, so you can see what each size costs you. The GitHub Actions billing guide shows how to read the report.

What a month costs on GitHub at 10,000 to 100,000 job-minutes

On GitHub Team, after the $18 of included compute, assuming it all goes to these jobs:

Each cell is the minutes × the rate − $18. Minutes here are billed minutes, so per-job rounding is already in them. On Enterprise Cloud, take $300 off instead of $18.

Same-size jobs at a flat price

Skipbox's plans run jobs of the same three sizes for a fixed monthly price with unlimited minutes and no usage charges. The flat price is lower once the same minutes would cost more on GitHub's runner of that size:

job-minutes a month above which the flat price is lower = (the flat price + the compute GitHub includes) ÷ GitHub's per-minute rate, rounded up to the hundred.

On GitHub Team, with $18 included, at GitHub's rates read on November 9, 2026:

Every volume in the GitHub table above is past these numbers. On Enterprise Cloud, put $300 in place of $18 in the formula; the cost calculator, signed in, prices your last 30 days on GitHub Free, Team and Enterprise Cloud.

A plan sets how many jobs run at once as well as how large each can be. Plan M's 8 vCPU run one skipbox-8x16 job at a time, or smaller jobs side by side while they fit; the rest wait in your account's queue, and GitHub cancels a job queued for 24 hours. When many jobs start together, or your minutes don't fit one job at a time, you add plans and the bill is their sum. The calculator works out the plans your jobs need.

Memory per core: where it isn't like for like

GitHub's larger runners have 4 GB of memory per core, against 8 GB for 2 cores on a private repository's ubuntu-latest (every standard runner's specs). Skipbox's VMs have 2 GB per vCPU:

For a job limited by CPU (compiling, most test suites), the same-size table holds. For a job limited by memory, size it by memory instead: a job that needs 16 GB runs on GitHub's 4-core larger runner, and on Skipbox it needs skipbox-8x16, plan M at $79 a month for 16 GB. The same-size break-even above doesn't apply to that pairing. A job that ends with exit code 137 has run out of memory; the exit code 137 guide covers how to tell and what to change.

What larger runners do that Skipbox doesn't

Those jobs stay on GitHub. Skipbox runs Linux x64 jobs from private repositories, each in a fresh Ubuntu 24.04 VM that is destroyed after the job.

Setting up a larger runner

1. On the organization's page, click Settings, then Actions and Runners in the sidebar.
2. Click New runner, then New GitHub-hosted runner.
3. Give it a name, and pick the platform, image, size, runner group and maximum concurrency, then click Create runner.
4. Add the repositories that may use it.
5. In the workflow, point the job at it. GitHub gives each larger runner a label that matches its name:

runs-on:
  labels: ubuntu-24.04-16core

Running the same sizes on Skipbox

GitHub has two kinds of runners, GitHub-hosted and self-hosted, and Skipbox's runners register with your account as self-hosted runners. GitHub bills nothing for self-hosted runner minutes and doesn't count them against your included compute, so a job you move doesn't spend from your Actions budget.

1. Sign in at skipbox.dev with GitHub and install the Skipbox CI app on your organization, or the Skipbox (personal) app on your own account.
2. Change one job's runs-on line to skipbox, skipbox-8x16 or skipbox-16x32 and push. A job larger than your largest plan is refused, and the reason shows on your account's page.
3. To go back, change the line back.

runs-on: skipbox (4 vCPU, 8 GB) runs on every plan, skipbox-8x16 needs plan M or L, and skipbox-16x32 needs plan L.

You can try it on your own jobs first: free for 7 days or 3,000 minutes, no card.

Only runs-on: skipbox jobs run in the free trial, one at a time, so a trial tests the 4 vCPU size and no larger.

Sources

All read on 2026-11-09.

• GitHub Docs, Actions runner pricing: every rate and SKU above, rounding up per job, larger runners only on Team and Enterprise Cloud, not free for public repositories.
• GitHub Docs, Larger runners reference: cores, memory and storage by size, the GPU and macOS sizes, static IP addresses on Enterprise Cloud and the limit of 10 pools, autoscaling to a maximum concurrency, the card and spending-limit requirements.
• GitHub Docs, Managing larger runners: organization owners add them, the setup steps.
• GitHub Docs, Using larger runners: the label that matches the runner's name, the runs-on examples.
• GitHub Docs, About Azure private networking for GitHub-hosted runners in your organization: larger runners only, 2 to 64 vCPU Ubuntu and Windows.
• GitHub Docs, GitHub Actions billing: included compute by plan in dollars, larger runners drawing on it on Team and Enterprise Cloud, self-hosted runners free.
• GitHub Docs, GitHub-hosted runners reference: ubuntu-latest on private repositories, 2 cores and 8 GB.
• GitHub Docs, Actions limits: a job queued for 24 hours is cancelled.

Updated November 9, 2026