An alternative to GitHub's larger runners

GitHub’s larger runners are its own machines with more CPUs, billed by the minute from the first minute. A job moves from one to a machine of the same size in your own cloud by changing its runs-on line.

jobs:
  build:
    # runs-on: ubuntu-24.04-16core
    runs-on: superci-16cpu

Which label for which runner

A larger runner has the name it was given when it was made. What matters is its size:

GitHub’s larger runner	GitHub, a minute	The label	AWS spot, a minute	AWS on-demand, a minute
Linux 4-core, 16 GB	$0.012	`superci-4cpu`	$0.0014	$0.0043
Linux 8-core, 32 GB	$0.022	`superci-8cpu`	$0.0030	$0.0083
Linux 16-core, 64 GB	$0.042	`superci-16cpu`	$0.0058	$0.016
Linux 32-core, 128 GB	$0.082	`superci-32cpu`	$0.010	$0.033
Linux 64-core, 256 GB	$0.162	`superci-64cpu`	$0.019	$0.065

Each label gets the same CPUs and memory as the runner beside it. AWS’s prices are for us-east-1 on 12 October 2026, with the machine’s 60 GB disk and public address counted in; spot prices move through the day. The calculator works out a month.

The other kinds have guides of their own: Arm runners, Windows runners and GPU runners.

What changes

Disk. GitHub’s larger runners come with 150 GB of disk and up. A SuperCI machine on AWS has 60 GB unless the label asks: superci-16cpu-300disk.
More than 32 CPUs. A label may ask for at most 32 CPUs until you raise that under Settings → Limits. AWS has machines up to 192.
Spot machines. Jobs run on AWS’s spare capacity unless the label says -ondemand. AWS can take a spot machine back; the job is then run again off spot. See Spot machines.
Starting. A machine takes about 15 seconds to start on AWS.
The plan you are on. GitHub offers larger runners on its Team and Enterprise Cloud plans. SuperCI works with any plan, and with GitLab.

What GitHub’s larger runners also do

On GitHub’s larger runners	With SuperCI
Static IP addresses	Start machines in your own subnets, without public addresses, behind a NAT gateway whose address is fixed.
Azure private networking	Your own subnets and security groups on AWS.
A limit on jobs at once	Jobs at once for each provider, and a monthly budget. See Order and limits.
Runner groups	The repositories you chose when you connected GitHub. There are no groups within them.
Custom images	No. Every AWS machine starts from the same image, with the tools of GitHub’s own Ubuntu 24.04 runners.
macOS	No. Keep macOS jobs on GitHub’s runners.

Moving a workflow

Start the dashboard and connect the repository.
Change runs-on on one job, push, and watch it under Jobs.
Change the rest. A workflow can use both: a job on macos-latest and a job on superci-16cpu run side by side.

If your tests do not use every CPU of a bigger machine, Faster tests says how to make them.

Sources

GitHub’s pages, read on 12 October 2026: Larger runners, Larger runners reference and Actions runner pricing.