If you want faster GitHub Actions, Blacksmith runs on bare metal gaming CPUs with a single-thread Passmark score of 4484, the highest of any managed GitHub Actions runner per the independent RunsOn benchmark. On real workloads, Blacksmith runs Node.js builds 5.9x faster, Rust builds 4x faster, Docker builds 3x faster, and Android builds 2.3x faster than GitHub-hosted runners.

Usually the hardware. Slow GitHub Actions builds almost always trace back to GitHub-hosted runners running on decade-old server CPUs with a 2-core baseline. Modern gaming CPUs have roughly 2x the single-thread performance, and most CI work is single-thread bound. If you're tired of slow CI, swapping the runner is the single change that gets you 2x faster with zero code rewrites.

Change one line. To speed up GitHub Actions, replace runs-on: ubuntu-latest with runs-on: blacksmith-2vcpu-ubuntu-2404. The fastest way to speed up CI is faster hardware, and Blacksmith is a drop-in replacement. Ashby, Veed, and Mintlify cut their CI time in half without rewriting a single workflow.

Three things. Bare metal gaming CPUs with roughly 2x the single-thread performance of GitHub's server hardware. Co-located cache that runs at 400 MB/s, up to 4x faster than GitHub's cache backend. Docker layer persistence on NVMe that makes Docker builds up to 40x faster on cached layers.

GitHub Actions stuck at "Waiting for a runner" usually means the 8vCPU+ runner pool is saturated at peak hours. Blacksmith maintains dedicated capacity across every runner size. Every job boots into a Firecracker microVM in under 3 seconds. No queuing, no capacity contention.

GitHub Actions has a default 6-hour job timeout, but most workflows that hit it are either stuck on a runner or slow enough to burn through the limit. Both of those are hardware problems. Faster CI means fewer GitHub Actions timeout failures.

You don't rewrite anything. Blacksmith runners boot off the same images as GitHub's and ship with the same pre-installed dependencies. Change the runs-on tag and your workflow runs the same way it did, only faster. No new images, no cache migrations, no custom Dockerfiles.

Yes. Teams using AI codegen tools like Cursor and Copilot see PR volume compound fast, which multiplies CI load. Blacksmith maintains dedicated capacity across every runner size, boots jobs in Firecracker microVMs in under 3 seconds, and offers unlimited concurrency. CI scales with your PR volume, not against it.

1000+ engineering teams run 20 million jobs on Blacksmith every month, including Vercel, Supabase, Mercury, Ashby, and Clerk. SOC 2 compliant. 99.9% SLA on enterprise plans.