Runners

Linux x64 GitHub Actions runners

Dedicated AMD and Intel vCPU runners from 2 to 64 cores, with NVMe scratch disks and a persistent cache on every job.

Specifications

vCPU: 2 – 64, dedicated
Memory: 4 GB per vCPU
Scratch disk: Local NVMe, 150 GB – 2 TB
Images: Ubuntu 22.04, 24.04
Cold start: ~3s
Cache: Unlimited, 1 GB/s restore

Runner labels

runnerhut-2vcpu-ubuntu-2404
runnerhut-4vcpu-ubuntu-2404
runnerhut-8vcpu-ubuntu-2404
runnerhut-16vcpu-ubuntu-2404
runnerhut-32vcpu-ubuntu-2404
runnerhut-64vcpu-ubuntu-2404

Standard GitHub-hosted Linux runners give you 2 shared vCPUs and a 14 GB SSD. Shared is the operative word: your build competes with whatever else landed on that host, which is why the same job takes 4 minutes on one run and 9 on the next.

runnerhut Linux runners are dedicated. The vCPUs are yours for the duration of the job, the scratch disk is local NVMe rather than network-attached, and the cache is restored over a 1 GB/s link instead of the throttled path actions/cache uses.

Switching

yaml

jobs:
  build:
-   runs-on: ubuntu-latest
+   runs-on: runnerhut-8vcpu-ubuntu-2404
    steps:
      - uses: actions/checkout@v4
      - run: make test

Change the label. Nothing else in the workflow moves.

Choosing a size

Bigger is not automatically faster. A test suite that never saturates 4 cores runs at the same wall-clock time on 16 and costs four times as much. Start one size up from your current runner, look at the CPU utilisation chart in the dashboard, and stop when the curve flattens.

Workload	Start with	Why
Node/TS unit tests	4 vCPU	Jest and Vitest parallelise well but are I/O bound
Go build + test	8 vCPU	The compiler scales nearly linearly
Rust release build	16 vCPU	codegen units keep every core busy
Docker image build	8 vCPU	Usually cache- and network-bound, not CPU-bound
Monorepo full CI	32 vCPU	Only if your task graph is genuinely parallel

Your next build could be twice as fast, at half the price

Start free. Migrating away is the same one line, and we publish that diff too.

Start building See pricing

Native arm64 GitHub Actions runnersNative Ampere and Graviton-class arm64 runners. No QEMU, no emulation penalty, 20% cheaper than the x64 equivalent.Local NVMe scratch disksEvery runner gets a local NVMe scratch disk instead of network-attached storage — the difference shows up in every checkout, install and test run.Preinstalled software on runnersThe full toolchain list on every runner image, and how it maps to the GitHub-hosted image so nothing breaks when you switch.Ubuntu versions on Linux runnersSupported Ubuntu images, the deprecation policy, and how to pin an image so a rollout never surprises your pipeline.