Short answer
In CI Bench v1 the 4 vCPU larger runner was 1.6x as fast as the 2 vCPU runner on median job time overall (faster on 11 of 11 single-job workloads) but its list price is twice as high ($0.012 against $0.006 a minute), so its cost per job was 39% higher than the 2 vCPU runner's (a ratio of 1.39). A larger runner pays off only when the job's time falls by more than the price ratio. Maintained by Vitko, Inc., which sells one of the products measured.
Workload by workload
| Workload | 2 vCPU median | 4 vCPU median | 4 vCPU speed vs 2 vCPU | 2 vCPU mean cost | 4 vCPU mean cost |
|---|---|---|---|---|---|
| cli | 368 s | 192 s | 1.9x faster | $0.039 | $0.045 |
| codex-lint | 884 s | 476 s | 1.9x faster | $0.087 | $0.099 |
| container | 239 s | 185 s | 1.3x faster | $0.027 | $0.042 |
| fd | 92 s | 66 s | 1.4x faster | $0.012 | $0.019 |
| flask | 23 s | 21 s | 1.1x faster | $0.0060 | $0.012 |
| guava | 1167 s | 946 s | 1.2x faster | $0.109 | $0.181 |
| monorepo | 384 s | 220 s | 1.7x faster | $0.040 | $0.046 |
| pydantic | 179 s | 154 s | 1.2x faster | $0.020 | $0.033 |
| ts | 784 s | 435 s | 1.8x faster | $0.077 | $0.099 |
| vite | 603 s | 288 s | 2.1x faster | $0.061 | $0.060 |
| vue | 121 s | 67 s | 1.8x faster | $0.014 | $0.020 |
List prices ($0.006 and $0.012 a minute, GitHub's runner pricing, read 2026-10-27); each job rounded up to a whole minute. A job that takes 100 s on 2 cores and 50 s on 4 cores costs the same on both only if billing were per second; with per-minute rounding check your own job lengths. List price per minute is not cost per job; faster runners finish sooner.
More: all questions answered · 12 options compared · ranked by speed · ranked by cost per job · burst workload.
Method: how the runs were made; data: results.json; every raw record: the repository. Maintained by Vitko, Inc., which sells one of the products measured. Page dated October 28, 2026.