machine.dev vs
GitHub-hosted runners.
GitHub's 4-vCPU Linux runner costs $0.012/min after the January 2026 price cut. machine.dev's 4-vCPU spot runner costs $0.00064/min — 18.8× cheaper. On-demand is still 2× cheaper. Same Ubuntu image, same Actions runner, one runs-on line changed.
The default.
GitHub's hosted runners are the path of least resistance. They work, they're maintained, and you don't have to think about them. If your builds are fast and your bill is fine, you don't need this page.
On January 1, 2026 GitHub cut hosted runner prices by ~39%. The CPU 4-core dropped from ~$0.020 to $0.012/min; the T4 GPU dropped from ~$0.07 to $0.052/min. That cut squeezed every “cheaper than GitHub” pitch in the market.
It didn't squeeze ours. machine.dev spot is 14–19× cheaper at every CPU size and 11–19× cheaper on T4. On-demand is still ~2× cheaper. If your builds are slow, your bill isn't fine, or you need a GPU bigger than a single T4 — read on.
Every size, every tier.
Per-minute USD. GitHub prices reflect their January 2026 cuts. machine.dev prices are pulled directly from the live API.
[NOTE_SWEEP] machine.dev wins on both spot and on-demand at every CPU size against GitHub-hosted. The spot delta is uniformly 14–19×. The on-demand delta is uniformly ~1.7–2×.
[NOTE_FREE] GitHub's included free minutes (2,000/mo Free, 3,000/mo Pro) cannot be used on larger runners or GPU runners. machine.dev's $10 free trial works on every size, including GPU.
One GPU vs Five.
GitHub publishes a single Tesla T4 SKU at $0.052/min. machine.dev offers T4G ARM, T4, L4, A10G, L40S, RTX 6000, plus AWS Trainium and Inferentia2.
Beyond price.
| SPEC | GITHUB_HOSTED | MACHINE_DEV |
|---|---|---|
| GPU SKU count | 1 (Tesla T4 only) | 6 NVIDIA tiers + Trainium / Inferentia2 |
| ARM GPU runners | No | Yes (T4G — only provider in market) |
| Spot tier | No | Yes (14-19× cheaper than OD) |
| CUDA pre-installed | No | Yes (12.1 + cuDNN 9.2.1, drivers 555.58) |
| Free tier on big runners | No (free min cap to ubuntu-latest) | Yes ($10 free compute, any size) |
| Data transfer fees | No | No (bundled) |
| Self-hosted platform fee | $0.002/min (paused, may return) | No |
| Setup | Zero (native) | One runs-on change |
The only line you touch.
Your workflow files, steps, secrets, caches, and Actions syntax stay exactly as they are. The only change is the runs-on label.
Three strengths.
Zero setup
GitHub runners are the default. No GitHub App to install, no labels to remember. If you're at low volume and your CPU bill is fine, you don't need this page.
Free minutes on public repos
GitHub gives unlimited Linux runner minutes on public repositories. machine.dev's free tier is $10 of compute, enough for ~6,000 minutes on a 2-vCPU spot — but it's a credit allowance, not a public-repo unlimited.
Tightest GitHub integration
Billing rolls into your existing GitHub invoice. Some GitHub Enterprise features (custom runner groups, network configs) are tied directly to GitHub-managed runners.
Five strengths.
94% cheaper on spot
Even after GitHub's January 2026 price cut, machine.dev's spot tier is 14-19× cheaper at every CPU size and 11-19× cheaper on T4 GPU. Same Ubuntu image, same Actions runner — different bill.
~50% cheaper on on-demand
Don't want to think about spot interruptions? On-demand machine.dev runners are still ~1.7-2× cheaper than the equivalent GitHub-hosted size, with no included-minutes carve-out for big runners or GPU.
Six GPU tiers vs one
GitHub offers a single Tesla T4 SKU at $0.052/min. machine.dev offers T4G (ARM), T4, L4, A10G, and L40S — plus AWS Trainium and Inferentia2 for inference. T4G ARM 8vCPU spot is $0.00277/min: cheaper than any GPU runner anywhere.
CUDA pre-installed
torch.cuda.is_available() returns True on the first run. No nvidia-smi setup steps, no apt-get installing 4 GB of CUDA libraries before your training loop starts.
No looming self-hosted tax
GitHub announced a $0.002/min platform fee for all self-hosted runners (ARC, RunsOn, Actuated, custom Terraform) targeted for March 2026. Paused after community backlash, but not cancelled. machine.dev absorbs the cost — your credit balance doesn't change if it returns.
Live per-minute rates
See what you'd actually pay. Compare with GitHub's $0.006/min for 2-core, $0.012/min for 4-core, and $0.052/min for GPU.
Get started free.
$10 free compute on signup. Enough to run real workloads and compare cost before committing.