Skip to content

Compare

Graphide CI or Blacksmith for your GitHub Actions jobs

Both keep your workflows on GitHub Actions and move jobs to their own runners by changing each job’s runs-on line. They differ in what they run, who tunes the jobs and how you pay. Here is what each company’s own pages say, so you can choose for your workflows.

Blacksmith’s side quotes its own pages, read on November 19, 2026. Graphide’s side quotes How Graphide CI works, CI security and Pricing.

Side by side

CriterionBlacksmithGraphide CI
What changes in your workflowEach job’s runs-on line changes to a blacksmith-* tag: “Transitioning to Blacksmith is as simple as changing a single line in your workflow file.” A Migration Wizard updates the workflow files. Overview · QuickstartYou press Switch to Graphide CI on a repository, or ask your coding agent to. Graphide opens one pull request that changes each moved job’s runs-on line to runs-on: graphide; your workflows stay on GitHub Actions. To move back, Switch back to GitHub Actions opens a pull request that puts each line back. /docs/ci/
What runs, and the result“Blacksmith runners boot off of the same image(s) as GitHub’s runners and have the exact same environment.” OverviewEvery test your job runs on GitHub runs on Graphide’s runners too, and a failing test always fails the job: Graphide CI changes how long you wait, never whether a job passes. /docs/ci/
Machine sizes2, 4, 8, 16 and 32 vCPUs for Ubuntu x64, Ubuntu ARM and Windows; 6 and 12 vCPUs for macOS. An 8 vCPU Ubuntu x64 runner has 32 GB of memory. OverviewOne label: “an x64 Linux machine with 8 vCPUs and 16 GiB”. Jobs that ask for more than 8 vCPUs keep running where they do today. /docs/ci/
PlatformsUbuntu x64 and ARM, Windows Server 2025 (public beta), and macOS “on Apple Silicon M4 chipsets”. On Windows, “Docker Linux containers are not supported”. Overviewx64 Linux: jobs on GitHub’s standard Ubuntu runners or on your own x64 Linux runners. macOS, Windows and Arm jobs, your own runners for another platform, deploy jobs (environment:) and npm provenance publishes keep running where they do today. /docs/ci/
Where jobs runLinux and Windows jobs run “in ephemeral Firecracker microVMs”, in us-east (Ashburn), us-west (Phoenix), eu-west (Amsterdam) or eu-central (Frankfurt). “Blacksmith has no concurrency limits.” Overview · ConcurrencyMicrosoft Azure, in its West US and West US 2 regions. One job per machine, destroyed when the job ends; an 8 vCPU runner is a virtual machine on an Azure VM that Graphide runs, which several customers’ machines share. /docs/security/ci/
Caching“All official GitHub and popular third-party cache actions transparently interact with our co-located cache”. Docker layer caching uses Blacksmith’s own build actions and is a paid add-on. Overview · Quickstart · PricingDependencies and build caches are already on the runner when the job starts. /docs/ci/
Who tunes the jobsYou pick each job’s runner size. Right-sizing “finds jobs on the wrong size and opens a PR to move them”. Run history, logs, SSH access, test analytics and CI analytics are in the product. OverviewGraphide’s agents tune how your jobs run (caches, machines, and test splits where a check shows they are safe), all on our end. /docs/ci/
How you payPublished per-minute rates by runner type, 3,000 free minutes a month, paid add-ons (Docker layer caching, sticky disks, static IPs) and an Enterprise plan. PricingA service at one fixed monthly price, quoted from what you run today. A free trial of 7 days or $10 of CI usage, whichever comes first; nothing is ever charged automatically. /pricing/
GitHub account“Blacksmith is limited to GitHub organizations and not available for personal repositories.” Its GitHub App goes on every repository that uses a blacksmith-* tag. Quickstart · Overview“Graphide CI needs a GitHub organization; its runners do not take jobs from a personal GitHub account.” /docs/ci/

Blacksmith’s own speed and cost claims. Blacksmith’s overview says it is “twice as fast as GitHub’s decade-old server hardware for most CI jobs” and that its runners cost less per minute than GitHub’s. Its pricing page says its savings figures are “based on GitHub’s per-minute rate and an estimated 2x faster runtime on Blacksmith.” Those are Blacksmith’s figures, on its assumptions; we have not tested them. This page states no speed or cost figure for Graphide CI against Blacksmith.

Where Blacksmith fits better

  • Your jobs run on macOS, Windows or Arm. Blacksmith documents runners for all three; Graphide CI moves x64 Linux jobs only.
  • A job needs more than 8 vCPUs. Blacksmith lists sizes up to 32 vCPUs; Graphide CI leaves such jobs where they run today.
  • You want published per-minute rates, and to choose each job’s runner size yourself.
  • You want run history, logs, SSH access and test and CI analytics in the same product, as Blacksmith documents them.
  • You want your jobs to run in Europe or the eastern US: Blacksmith names regions in both.

Where Graphide CI fits better

  • You would rather not tune runners yourself: Graphide’s agents tune each job’s caches and machine on our end, and split a test step only where a check shows it is safe.
  • You want dependencies and build caches already on the runner when each job starts.
  • Your jobs are x64 Linux and fit in 8 vCPUs and 16 GiB.
  • You want one fixed monthly price, quoted from your usage, after a trial that never charges automatically.

How to measure it yourself

Neither company’s numbers tell you how your jobs will run. Run your own workload on each runner and time it against a stated baseline.

  1. Pick the workloads

    Choose the x64 Linux jobs you wait on most. Those are the jobs both can run, since Graphide CI moves only x64 Linux jobs.

  2. State the baseline

    Time the same jobs at the same commit on the runners you use today, and write down their size. Compare like with like: at 8 vCPUs, Blacksmith lists 32 GB of memory and Graphide CI’s runner has 16 GiB.

  3. Measure push to verdict

    Count from the push until the check reports pass or fail, queue time included, over several runs of each job. Use the median, not the best run.

  4. Check the verdict

    Compare test names and results in each runner’s log. A faster run that tested less is not the same job.

  5. Read it per workload

    A speedup on one job says nothing about another. Report each job’s result on its own, never as one number for all your CI.

Next steps

  • How Graphide CI works: what moves, what stays, and how to switch a repository and back.
  • Pricing: how the fixed monthly quote and the free trial work.
  • Graphide review is a separate product that checks each pull request with evidence. It works on its own, with or without Graphide CI.

Pages we read

Blacksmith’s pages change. Check them before you decide; where they differ from this page, they are right.

  1. Blacksmith docs: Instance Types (runner overview)

    https://docs.blacksmith.sh/blacksmith-runners/overview · read on November 19, 2026

  2. Blacksmith docs: Unlimited concurrency

    https://docs.blacksmith.sh/blacksmith-runners/concurrency · read on November 19, 2026

  3. Blacksmith docs: Quickstart

    https://docs.blacksmith.sh/introduction/quickstart · read on November 19, 2026

  4. Blacksmith: Pricing

    https://www.blacksmith.sh/pricing · read on November 19, 2026

  5. How Graphide CI works, CI security and Pricing

    Graphide’s own docs and pricing page, for every Graphide CI cell.