self-host →

//github actions cost calculator · aws spot pricing

what are your CI minutes actually costing you?

Punch in your monthly GitHub Actions minutes and see the bill on GitHub-hosted runners next to the same workload on self-hosted RunsOn — real EC2 spot prices in your own AWS account, no per-minute markup.

toggle a platform off by setting its minutes to 0

start from a common setup

Not sure where to begin? Load a typical CI shape and tweak from there — the calculator updates instantly.

per-runner cost, by size

GitHub-hosted vs RunsOn $/min for every runner size — across architectures and instance generations. Pick spot or on-demand and set your gp3 root volume to match your setup.

arch
generation
pricing
gp3 root gb mb/s
runner instance RunsOn / min GitHub / min you save RunsOn CPU GitHub CPU CPU vs GitHub
1cpu m7a.medium $0.0006 $0.0020 2,882 2,269 +27%
2cpu m7i-flex.large $0.0010 $0.0060 3,074 2,269 +35%
4cpu m7i-flex.xlarge $0.0017 $0.0120 3,074 2,269 +35%
8cpu c7i-flex.2xlarge $0.0027 $0.0220 3,038 2,269 +34%
16cpu c7i-flex.4xlarge $0.0043 $0.0420 10× 3,038 2,269 +34%
32cpu c7i-flex.8xlarge $0.0092 $0.0820 3,038 2,269 +34%
48cpu c7i-flex.12xlarge $0.0088 3,038
64cpu c7i-flex.16xlarge $0.0133 $0.1620 12× 3,038 2,269 +34%
96cpu c7i.24xlarge $0.0223 $0.2520 11× 3,010 2,269 +33%
spot prices · us-east-1 · incl. 30GB gp3 @ 400 MB/s billed by AWS, to the second

how this is calculated

01Where do the prices come from?+
RunsOn figures are real EC2 prices in us-east-1, including a 30 GB gp3 root volume — spot by default, with an on-demand toggle. GitHub figures are GitHub-hosted per-minute rates. Both are normalized to $/min and multiplied by your minutes. See the full runner cost table.
02Does the estimate include the RunsOn license?+
Yes. The yearly total adds the flat Commercial license ($350/yr) on top of compute, so the savings shown are net of what you pay RunsOn. Compute itself is billed by AWS, directly to your account — there is no per-minute markup.
03Why is self-hosted so much cheaper?+
You pay AWS spot prices for the exact instance a job needs, only while it runs, instead of a fixed per-minute rate with a margin baked in. On Linux that routinely lands 6–12× under GitHub-hosted. Faster CPUs also mean jobs finish sooner, which compounds the saving.
04What about caching, data transfer, and queue time?+
This calculator covers compute, which dominates the bill. RunsOn's S3-backed cache lives in your account (no egress to a third party), and faster runners cut wall-clock time further. Storage and transfer are typically cents per job.