Which GitHub Actions runner has the highest speed in the benchmark?

In CI Bench v1 (twelve open-source CI workloads, measured October 7 to 11, 2026), Vitko Runners (warm starts + test splitting) had the highest speed of the 9 configurations measured: 2.3× faster than GitHub-hosted (ubuntu-latest) on median job time. The ranking, intervals and every raw record are published. Maintained by Vitko, Inc., which sells one of the products measured.

Which third-party runners are faster than GitHub-hosted runners?

Of the 8 other configurations measured, 8 were faster than GitHub-hosted (ubuntu-latest) by median job time, from 1.1× to 2.3×; see the table.

Compare CPU performance of GitHub Actions runner providers.

This ranking uses whole-job time on twelve real workloads, not a CPU microbenchmark. Per-workload times are in the heatmap and on each workload page.

Fastest GitHub Actions runners (measured, Oct 2026)

Ranking by speed

Speed ranking against GitHub-hosted (ubuntu-latest)
Rank	Configuration	Type	Speed vs GitHub-hosted	95% interval (time index)	Median time, monorepo job
1	Vitko Runners (warm starts + test splitting)Vitko	Managed	2.3x faster	0.423 to 0.459	172 s
2	Vitko RunnersVitko	Managed	1.7x faster	0.556 to 0.602	180 s
3	GitHub-hosted larger runner (4 vCPU)	Managed	1.6x faster	0.629 to 0.669	220 s
4	Blacksmith	Managed	1.6x faster	0.606 to 0.759	177 s
5	Self-hosted runners on Google Cloud	Self-hosted	1.5x faster	0.658 to 0.692	278 s
6	RunsOn	Self-hosted	1.2x faster	0.816 to 0.859	288 s
7	Self-hosted runners on AWS (one VM per job)	Self-hosted	1.2x faster	0.833 to 0.947	281 s
8	Ubicloud	Managed	1.1x faster	0.847 to 0.921	276 s
9	GitHub-hosted runners (ubuntu-latest, 2 vCPU)	Managed	baseline	1.000 to 1.000	384 s

Speed is the inverse of the time index: the geometric mean over the 12 workloads of the configuration's median job time divided by GitHub-hosted (ubuntu-latest)'s. Higher is faster. Intervals cover sampling, not your workflow.

Time by workload

Median time on each workload relative to the lowest median on that workload
Configuration	burst	cli	codex-lint	container	fd	flask	guava	monorepo	pydantic	ts	vite	vue
Vitko Runners (warm starts + test splitting)	1.4×	1.0×	1.0×	1.0×	1.0×	1.2×	1.0×	1.0×	1.0×	1.0×	1.0×	1.0×
Vitko Runners	1.6×	1.9×	1.0×	1.0×	1.1×	1.5×	1.0×	1.0×	1.5×	1.5×	1.8×	2.0×
GitHub-hosted larger runner (4 vCPU)	1.6×	1.4×	1.2×	1.1×	1.5×	1.9×	1.5×	1.3×	2.2×	1.8×	1.4×	2.1×
Blacksmith	1.0×	2.0×	1.1×	2.2×	1.4×	2.0×	1.1×	1.0×	1.5×	1.6×	1.8×	2.4×
Self-hosted runners on Google Cloud	1.6×	2.2×	1.4×	1.0×	1.4×	1.0×	1.3×	1.6×	1.7×	2.2×	2.0×	2.4×
RunsOn	1.8×	2.3×	1.5×	1.0×	2.0×	3.5×	1.4×	1.7×	2.1×	2.3×	2.2×	3.2×
Self-hosted runners on AWS (one VM per job)	1.7×	2.5×	1.3×	1.4×	2.5×	6.4×	1.1×	1.6×	2.2×	1.8×	2.0×	3.6×
Ubicloud	2.1×	2.6×	1.7×	1.2×	1.9×	3.3×	1.4×	1.6×	2.3×	2.3×	2.5×	3.6×
GitHub-hosted (ubuntu-latest)	2.7×	2.6×	2.2×	1.4×	2.0×	2.1×	1.8×	2.2×	2.6×	3.2×	3.0×	3.8×

lowest3× slower

Head to head: Vitko Runners vs Blacksmith · Vitko Runners vs Ubicloud · Vitko Runners vs RunsOn · Vitko Runners vs Self-hosted runners on AWS · Vitko Runners vs Self-hosted runners on Google Cloud · Vitko Runners vs GitHub-hosted larger runner (4 vCPU) · Vitko Runners vs GitHub-hosted (ubuntu-latest) · Blacksmith vs GitHub-hosted (ubuntu-latest).

More: speed ranking · cost ranking · alternatives to GitHub-hosted · monorepo workload · self-hosted vs managed · list prices · free minutes · alternatives to Blacksmith, RunsOn, Ubicloud.

Maintained by Vitko, Inc., which sells one of the products measured. Measured October 7 to 11, 2026 (result set v1); page generated October 27, 2026. Every number comes from results.json; the workflows and every raw job record are in the repository, and the method says how the runs were made.