Blacksmith vs GitHub-hosted runners
Blacksmith suits teams on GitHub Actions that want microVM-isolated runners on bare metal with built-in Docker layer caching and CI observability. GitHub-hosted runners suits teams on GitHub Actions that want runners with no setup, and public repositories that can use the free standard runners. Blacksmith is a managed service hosted by the vendor for GitHub Actions, priced at $0.004/min; GitHub-hosted runners is the CI platform’s own hosted runners for GitHub Actions, priced at $0.006/min.
Key differences between Blacksmith and GitHub-hosted runners
- Type: Blacksmith is a managed service hosted by the vendor; GitHub-hosted runners is the CI platform’s own hosted runners.
- Isolation: Blacksmith: Firecracker microVM. GitHub-hosted runners: Fresh VM.
- Published price: $0.004/min (Blacksmith) versus $0.006/min (GitHub-hosted runners).
- Largest runner: 32 vCPU / 128 GB (Blacksmith) versus 96 vCPU / 384 GB (GitHub-hosted runners).
What they have in common
- Both companies are based in the United States.
- Both run GitHub Actions jobs.
- Both document Linux, macOS and Windows runners.
- Both document arm64 runners.
- Both give every job a fresh environment.
- Neither is open source.
Blacksmith vs GitHub-hosted runners at a glance
| Attribute | Blacksmith | GitHub-hosted runners |
|---|---|---|
| What it is | Managed GitHub Actions runners in Firecracker microVMs on the vendor’s own bare-metal fleet, with caches kept next to the runners. | GitHub’s own Actions runners: a new VM per job on Linux, Windows and macOS, with larger, arm64 and GPU sizes billed per minute. |
| Type | Managed runner service | First-party hosted runners |
| CI systems | GitHub Actions | GitHub Actions |
| Operating systems | Linux, macOS, Windows | Linux, macOS, Windows |
| Architectures | x64, arm64 | x64, arm64 |
| Isolation | Each Linux and Windows job runs in an ephemeral Firecracker microVM on Blacksmith’s bare-metal hosts and is destroyed afterwards; the isolation used for macOS runners is not documented | Each job runs in a new virtual machine hosted by GitHub; the 1-vCPU ubuntu-slim runner instead runs in an unprivileged container on a shared VM |
| Fresh environment per job | Yes | Yes |
| Hardware | Blacksmith describes its hosts as bare-metal machines with gaming CPUs chosen for single-core performance and NVMe drives, in ISO 27001 data centers; CPU models are not named, and macOS runners use Apple M4. | Azure virtual machines, with eligible 2-core and 4-core Linux jobs also scheduled on AWS; standard macOS runners are 3-core M1 or 4-core Intel, and the macOS XLarge runner is a 5-core M2 |
| Performance (vendor claim) | Blacksmith says most CI jobs run about 2x faster than on GitHub-hosted runners and cache downloads about 4x faster, and that customers report 2x to 40x faster Docker builds (vendor claims) | Not documented |
| Runner sizes | 2 to 32 vCPU with 8 to 128 GB RAM on Linux x64; 2 to 32 vCPU with 6 to 96 GB on Linux arm64; 2 to 32 vCPU with 7 to 112 GB on Windows x64 (public beta); 6 or 12 vCPU with 24 or 48 GB on macOS | Standard: 1 vCPU / 5 GB (ubuntu-slim); 2 vCPU / 8 GB in private and 4 vCPU / 16 GB in public repositories on Linux and Windows, x64 and arm64. Larger runners: 2 to 96 vCPU on x64 and 2 to 64 vCPU on arm64; macOS 3 to 12 cores |
| Caching | Standard cache actions (actions/cache and the setup-* actions) are served from a cache in the same data center as the runners, with 25 GB of free storage per repository per week and eviction of entries unused for 7 days. Docker layers persist on NVMe-backed sticky disks through useblacksmith/setup-docker-builder, billed at $0.50/GB/month; the docs also cover Git checkout, container image and Bazel, Go, Turborepo and Nx build caching. | Dependency caching through the cache action, with 10 GB per repository included; entries not accessed for 7 days are removed and the oldest are evicted at the limit. The limit can be raised, billed at $0.07 per GB per month. Larger runners can use custom VM images. |
| Docker layer caching | Yes | Not documented |
| GPU | Not documented | GPU larger runners have 4 vCPU, 28 GB RAM and one Tesla T4 with 16 GB VRAM, at $0.052/min on Linux and $0.102/min on Windows. |
| Concurrency | Blacksmith documents no limit on concurrent jobs or vCPUs | Standard runners: 20 concurrent jobs on Free, 40 on Pro, 60 on Team and 500 on Enterprise, of which 5 macOS (50 on Enterprise). Larger runners: up to 1,000 concurrent jobs and 100 GPU jobs |
| Static IPs | Yes | Yes |
| SSH into running jobs | Yes | Not documented |
| Pricing | The pricing page lists Ubuntu x64 at $0.004/min, Ubuntu ARM at $0.0025/min, Windows x64 at $0.008/min and macOS M4 at $0.08/min; the docs say runners cost half of GitHub’s per-minute rate. Add-ons: Docker layer caching and sticky disks at $0.50/GB/month and static IPs at $100/IP/month. Flex scheduling costs 40% less for jobs that can wait up to 24 hours to start. | Private repositories, per minute: Linux 2-core x64 $0.006, Linux 2-core arm64 $0.005, Linux 1-core $0.002, Windows 2-core $0.010, macOS 3- or 4-core $0.062. Larger Linux x64 runners run from $0.012 (4-core) to $0.252 (96-core); larger macOS runners are $0.077 (12-core Intel) and $0.102 (M2 Pro). Usage is rounded up to the whole minute. |
| Free tier | 3,000 free minutes per organization per month, counted as x64 2-vCPU minutes; larger, ARM, Windows and macOS runners consume them at weighted rates | Standard runners are free in public repositories. Private repositories get $12 of included compute usage per month on GitHub Free, $18 on Pro and Team, and $300 on Enterprise Cloud |
| Open source | No | No |
| Self-host / your cloud | No / Not documented | No / No |
| Compliance | SOC 2 Type 2; GDPR compliant (as stated by Blacksmith); HIPAA described as in progress | Not documented |
| Regions | us-east (Ashburn), us-west (Phoenix), eu-west (Amsterdam) and eu-central (Frankfurt); changing region requires contacting support | Regions are not documented for github.com. GitHub says Windows runners and other Linux runners run on Microsoft Azure, eligible Linux jobs on Azure or AWS, and macOS runners in Azure data centers |
| Company headquarters | San Francisco, California, United States | San Francisco, California, United States |
| Good fit for | Teams on GitHub Actions that want microVM-isolated runners on bare metal with built-in Docker layer caching and CI observability | Teams on GitHub Actions that want runners with no setup, and public repositories that can use the free standard runners |
Verified November 22, 2026 from each provider’s documentation. Performance figures are vendor claims, not our measurements.
Which should you choose?
Reasons to choose Blacksmith
- You want each Linux or Windows job in its own Firecracker microVM that is destroyed after the job
- You want Docker layers kept on persistent NVMe-backed disks between runs
- You need to run many jobs at once: Blacksmith documents no concurrency limit
- You want to SSH into running jobs and search CI logs from one console
Reasons to choose GitHub-hosted runners
- You want runners that need no installation or account beyond GitHub itself
- You maintain public repositories, where standard runners are free
- You need Linux, Windows and macOS, on x64 and arm64, from one provider
- You want every job to start in a clean VM that is not reused
Trade-offs to check first
- Blacksmith: Available to GitHub organizations only, not personal repositories
- Blacksmith: Windows runners are in public beta and do not support Docker Linux containers
- GitHub-hosted runners: Larger runners are billed in public repositories and need a GitHub Team or Enterprise Cloud plan
- GitHub-hosted runners: Jobs are limited to 6 hours of execution time
Run your slowest workflow on both before deciding. See the full Blacksmith profile and GitHub-hosted runners profile, or the complete comparison.
More comparisons
- Blacksmith vs Depot GitHub Actions Runners
- Blacksmith vs Luxor Runners
- Blacksmith vs Namespace
- Blacksmith vs Ubicloud Runners
- Blacksmith vs WarpBuild
- Avrea vs Blacksmith
- Blacksmith vs BuildPulse Runners
- Blacksmith vs CRACI
- Actions Runner Controller (ARC) vs GitHub-hosted runners
- Depot GitHub Actions Runners vs GitHub-hosted runners
- GitHub-hosted runners vs GitLab-hosted runners
- GitHub-hosted runners vs Namespace
- GitHub-hosted runners vs Ubicloud Runners
- GitHub-hosted runners vs WarpBuild
Frequently asked questions
Is Blacksmith or GitHub-hosted runners better?
What is the difference between Blacksmith and GitHub-hosted runners?
Which is open source, Blacksmith or GitHub-hosted runners?
How do Blacksmith and GitHub-hosted runners isolate jobs?
How does Blacksmith pricing compare with GitHub-hosted runners?
Is Blacksmith or GitHub-hosted runners faster?
Do Blacksmith and GitHub-hosted runners cache Docker layers?
Do Blacksmith and GitHub-hosted runners offer GPU runners?
Where do Blacksmith and GitHub-hosted runners runners run?
Sources
- blacksmith.sh/
- blacksmith.sh/pricing
- blacksmith.sh/security
- docs.blacksmith.sh/blacksmith-runners/overview
- docs.github.com/en/billing/reference/actions-runner-pricing
- docs.github.com/en/billing/concepts/product-billing/github-actions
- docs.github.com/en/actions/reference/runners/github-hosted-runners
- docs.github.com/en/actions/reference/runners/larger-runners
Check before you rely on it. Products change quickly, and this page reflects public documentation as of November 22, 2026. Confirm anything that matters in the provider’s own documentation.
Cite this page
CI Runners. (2026). Blacksmith vs GitHub-hosted runners. cirunners.org. Last reviewed November 22, 2026. https://cirunners.org/compare/blacksmith-vs-github-hosted-runners/