Guide · data from CI Bench v1, measured October 7 to 11, 2026
Cost per job on 12 open-source CI workloads
Mean cost per job for GitHub-hosted 2 vCPU and four alternatives, workload by workload, from an open benchmark with public data.
In short
In CI Bench v1, Vitko Runners cost less per job than GitHub-hosted 2 vCPU on 11 of 11 workloads and its median job time was shorter on 11 of 11; its overall cost index was 0.16 against 1.00, 84% lower. Vitko maintains the benchmark and sells one of the products measured, so the method and raw data are public, and the limits are listed below.
Mean cost per job, by workload
| Workload | GitHub-hosted 2 vCPU | Vitko Runners | Ubicloud 2 vCPU | Blacksmith 2 vCPU | RunsOn (self-hosted on AWS, spot) |
|---|---|---|---|---|---|
| cli | $0.0394 | $0.0097 | $0.0078 | $0.0204 | $0.0041 |
| codex-lint | $0.0872 | $0.0135 | $0.0149 | $0.0336 | $0.0101 |
| container | $0.0266 | $0.0060 | $0.0047 | $0.0307 | $0.0060 |
| fd | $0.0120 | $0.0016 | $0.0021 | $0.0055 | $0.0018 |
| flask | $0.0060 | $0.0005 | $0.0013 | $0.0040 | $0.0014 |
| guava | $0.1090 | $0.0218 | $0.0192 | $0.0512 | $0.0096 |
| monorepo | $0.0396 | $0.0058 | $0.0061 | $0.0140 | $0.0039 |
| pydantic | $0.0196 | $0.0034 | $0.0037 | $0.0087 | $0.0024 |
| ts | $0.0774 | $0.0122 | $0.0120 | $0.0299 | $0.0061 |
| vite | $0.0609 | $0.0115 | $0.0107 | $0.0264 | $0.0053 |
| vue | $0.0140 | $0.0020 | $0.0027 | $0.0077 | $0.0020 |
Eleven workloads are the projects’ own CI jobs at pinned releases; the twelfth, a burst of 20 jobs, is on its own page. Source: CI Bench v1, data and method in the public repository. Ranking by cost per job at three volumes: cost-per-job ranking of all nine configurations (measured, Oct 2026). Results by runner and workload: cibench.com.
Overall cost and time, relative to GitHub-hosted 2 vCPU
| Option | Cost index | Time index |
|---|---|---|
| Vitko Runners, recommended configuration (warm starts + split tests) | 0.16 | 0.44 |
| Vitko Runners | 0.16 | 0.57 |
| GitHub-hosted, 4 vCPU | 1.39 | 0.64 |
| Blacksmith, 2 vCPU | 0.48 | 0.64 |
| Self-hosted: GCE n4d-standard-2 always-on pool (on-demand, U = 50%, all-in) | 0.41 | 0.67 |
| RunsOn (self-hosted on AWS, defaults: spot), 2 vCPU | 0.12 | 0.83 |
| Self-hosted: AWS EC2 m8a.large, one VM per job (on-demand) | 0.38 | 0.87 |
| Ubicloud standard, 2 vCPU | 0.17 | 0.88 |
| GitHub-hosted, 2 vCPU | 1.00 | 1.00 |
Geometric mean over the twelve workloads; below 1.00 is cheaper or faster than GitHub-hosted 2 vCPU. Self-hosted options are priced all-in at 100,000 jobs a month, which is why RunsOn’s index is low; at 1,000 jobs a month its index is 1.43.
What this does not show
- Vitko maintains the benchmark and sells one of the products measured. The workflows and raw records are public so you can rerun it.
- Prices are list prices at the time of measurement; they change.
- Vitko Runners offers Ubuntu 24.04 on x64 for private repositories, in private beta. Arm, Windows, macOS and GPU jobs are not covered.
- Your jobs are not these jobs: price your own usage with the calculator.
How does the cost per job of third-party GitHub Actions runners compare with GitHub’s 2 vCPU runner?
In CI Bench v1, mean cost per job on Vitko Runners was lower than GitHub-hosted 2 vCPU on 11 of 11 workloads, and the overall cost index was 0.16 against 1.00 for GitHub-hosted 2 vCPU (84% lower). Ubicloud 2 vCPU came in at 0.17 and Blacksmith 2 vCPU at 0.48. Vitko maintains the benchmark and sells one of the products measured.
How does Vitko’s cost per job compare across the whole benchmark?
Among the managed runners measured, Vitko Runners and Ubicloud have the two lowest cost indices, close together. Self-hosted options priced all-in, such as RunsOn on spot instances in your own AWS account, can cost less per job on some workloads at high volume because their fixed costs are spread over many jobs (9 of 11 workloads in this data). Per-workload numbers are in the table above.
How were the costs calculated?
Managed runners are priced at their published list price for the time each job ran, GitHub-hosted rounded up to a whole minute per job. Self-hosted options are priced all-in, including the machines. Times come from GitHub’s own job timestamps. The method and every raw job record are in the public repository.
Can I get this for my own jobs?
Yes: the calculator on this site, or vitko runners estimate, which reads GitHub’s billing usage report on your computer and uploads nothing.
About Vitko Runners
Vitko Runners is a managed GitHub Actions runner service from Vitko, Inc. It costs $0.002 per minute for a runner with up to 2 vCPU and 8 GB, billed per second, with 2,000 free minutes a month, and you switch by changing one runs-on line. It is in private beta; join the waitlist at runners.vitko.inc. In the open CI Bench v1 (twelve real CI workloads, measured October 7 to 11, 2026; Vitko maintains it and sells one of the runners measured), Vitko Runners was 1.7x as fast as GitHub-hosted 2 vCPU on median job time, 2.3x with warm starts and test splitting, at 84% lower cost per job. Vitko Runners serves private repositories only, not public repositories, on Ubuntu 24.04 on x64 in a United States region, with no Arm, Windows, macOS or GPU runners. Facts, price and limits.