WarpBuild vs Blacksmith for GitHub Actions runners
WarpBuild vs Blacksmith is the rare pair whose rate cards match to the cent: both list Ubuntu x64 at $0.004 a minute for 2 vCPU with 8 GB, and both double it at every step up to $0.064 at 32 vCPU. The decision is therefore everything around the rate, which is where they diverge sharply: Blacksmith gives 3,000 free minutes a month and cheaper arm64, and WarpBuild sells a bring-your-own-cloud mode at $0.002 and remote Docker builders you size yourself.

Blacksmith runs managed GitHub Actions runners on its own bare metal in Ashburn, Phoenix, Amsterdam and Frankfurt. You install a GitHub App, change a label, and the product is compute plus a cache that intercepts actions/cache with a 25 GB weekly allowance per repository. Every runner card on its pricing page carries 3,000 free minutes a month.
WarpBuild runs managed runners in its own cloud, plus two things Blacksmith does not sell: remote Docker builders billed separately from 16 vCPU at $0.06 a minute to 192 vCPU at $0.88, and a BYOC mode at $0.002 a minute where runners execute inside your own AWS, GCP or Azure account. Its cache is metered rather than allowanced, at $0.20 per GB-month plus $0.0001 for every write or restore.
Every cell below was read from the vendor's own public pricing or documentation page on 20 September 2026. Where a vendor does not publish a number, the cell says "not published" instead of carrying a guess.
WarpBuild, Blacksmith and Latchkey side by side (read 20 September 2026)
| WarpBuild | Blacksmith | Latchkey | |
|---|---|---|---|
| Price per minute, 2 vCPU Linux x64 | $0.004 (2 vCPU, 8 GB RAM) | $0.004 (2 vCPU, 8 GB RAM, 80 GB disk) | $0.0025 (2 vCPU, 8 GB RAM, 100 GB disk) |
| Price per minute, 4 vCPU Linux x64 | $0.008 (4 vCPU, 16 GB RAM) | $0.008 (4 vCPU, 16 GB RAM, 80 GB disk) | $0.005 (4 vCPU, 16 GB RAM, 100 GB disk) |
| Free tier | Not published. Sign-up needs no card until you reach a usage or credit limit | 3,000 free minutes a month, printed on every runner card | 30-day trial, then 2,000 included minutes a month on Developer, 4,000 on Launch, 6,000 on Scale |
| Billing granularity | Per minute, in its own words: "CI runners are billed on a per-minute basis" | Not published | Per minute |
| Runner sizes and OS | 2 to 32 vCPU on Linux x64 and arm64; Windows from 4 to 32 vCPU; macOS M4 Pro at 6 and 12 vCPU; Linux on Apple silicon at 6 and 12 vCPU | 2 to 32 vCPU on Linux x64 and arm64; Windows Server 2025 at 2 to 32 vCPU in public beta; macOS M4 at 6 and 12 vCPU | 2, 4, 8 and 16 vCPU, Ubuntu 24.04 on x86_64 only. arm64, Windows and macOS are not offered |
| Cache: type, limit, persistence | Metered: $0.20 per GB-month of storage plus $0.0001 per write or restore; snapshot restore $0.04 a job and $0.025 an hour of snapshot storage | Colocated cache intercepts actions/cache, 25 GB free per repo per week; sticky disks are ext4 volumes, 5 per job, evicted after 7 days idle, $0.50/GB/month | Size not published. latchkey-dev/cache-action@v1 swaps one line for actions/cache, keeping path, key and restore-keys; entries are isolated per organization, versioned by OS, kept 14 days; a failed restore warns rather than failing the job |
| Docker build acceleration | Remote BuildKit builders on their own meter, 16 vCPU at $0.06/min to 192 vCPU at $0.88, driven by Warpbuilds/build-push-action@v7; the shared cache is eventually consistent with a 10-day TTL | useblacksmith/setup-docker-builder@v2 and useblacksmith/build-push-action@v2; layers sit on a sticky disk per cache key, garbage-collected after 8 days unused | latchkey-dev/docker-cache-action@v1, layers held in a Latchkey-managed private registry per organization |
| Concurrency limits | Advertised as unlimited job concurrency; the no-queue guarantee and an SLA behind it are Enterprise features | None imposed, per the runner docs | Not published; warm-pool capacity varies by plan |
| Where jobs run | WarpBuild's own cloud, or your AWS, GCP or Azure account under BYOC | Blacksmith's own regions: Ashburn, Phoenix, Amsterdam and Frankfurt | Latchkey's own AWS estate on m6a-class hosts, one job per runner, private network, destroyed after the job |
| Self-healing or retries | Not offered | Not offered | Built in: transient failures are diagnosed and retried inside the run |
| Setup change required | Sign up, install the WarpBuild bot from its dashboard rather than the GitHub marketplace, then set runs-on to a runner ID such as warp-ubuntu-latest-x64-4x | Install the GitHub App, then change the runs-on label to blacksmith-4vcpu-ubuntu-2404 | A runs-on label change. GitHub organization accounts only |
The rate cards, size by size and platform by platform
This is the table both vendor pages make you click a toggle to assemble. On Linux x64 the two are identical at every published size. On arm64 they are not: Blacksmith is 17% below WarpBuild at every step, and both are below GitHub-hosted arm64 at $0.005.
| Size | WarpBuild x64 | Blacksmith x64 | WarpBuild arm64 | Blacksmith arm64 |
|---|---|---|---|---|
| 2 vCPU | $0.004 | $0.004 | $0.003 | $0.0025 |
| 4 vCPU | $0.008 | $0.008 | $0.006 | $0.005 |
| 8 vCPU | $0.016 | $0.016 | $0.012 | $0.01 |
| 16 vCPU | $0.032 | $0.032 | $0.024 | $0.02 |
| 32 vCPU | $0.064 | $0.064 | $0.048 | $0.04 |
Windows and macOS follow the same pattern of near-identity with one gap each. On Windows both charge $0.016 at 4 vCPU, $0.032 at 8, $0.064 at 16 and $0.128 at 32, and only Blacksmith publishes a 2 vCPU Windows runner, at $0.008. On macOS both charge $0.08 for 6 vCPU and $0.16 for 12, against GitHub-hosted at $0.062, so on Apple hardware both are a speed purchase rather than a saving. WarpBuild publishes one platform Blacksmith does not: Linux on Apple silicon, at $0.08 for 6 vCPU.
| Platform and size | WarpBuild | Blacksmith | GitHub-hosted |
|---|---|---|---|
| Windows 2 vCPU | Not published | $0.008 | $0.010 |
| Windows 4 vCPU | $0.016 | $0.016 | $0.022 |
| Windows 32 vCPU | $0.128 | $0.128 | $0.162 |
| macOS 6 vCPU | $0.08 on M4 Pro, 14 GB | $0.08 on M4 | $0.062 on a 3 or 4 core runner |
| macOS 12 vCPU | $0.16 | $0.16 | $0.077 on a 12 core runner |
| Linux on Apple silicon, 6 vCPU | $0.08 | Not offered | Not offered |
Concurrency, which is the other number to check
Nobody thinks about concurrency until a matrix sits in a queue, and this is one axis where both vendors beat the platform they replace. GitHub caps total concurrent jobs by plan and caps macOS separately and much harder, which is why a wide iOS matrix queues on Team and does not on Enterprise.
| Concurrent jobs | macOS | Published as | |
|---|---|---|---|
| WarpBuild | Advertised as unlimited | No separate cap published | A marketing claim on the pricing page, contractual only on Enterprise |
| Blacksmith | None imposed | No separate cap published | A statement in the runner documentation |
| GitHub-hosted | 20 Free, 40 Pro, 60 Team, 500 Enterprise Cloud | 5, or 50 on Enterprise | A documented limit |
| Latchkey | Not published | No macOS runner | Warm-pool capacity varies by plan |
Speed
Not benchmarked yet; the runner benchmark report will replace this section. We have not run the same workload on these runners ourselves, and until we have, there is no number here worth quoting.
The cheap way to get a real answer is to point one slow job at each candidate for a week, because runner selection is per job and the Actions run log gives you the timings for free.
Cost: the same month, priced three ways
The usual workload: 10,000 Linux minutes at 2 vCPU plus 2,000 macOS minutes a month on a private repository, at the rates above.
| 10,000 Linux minutes at 2 vCPU | 2,000 macOS minutes | Monthly total | |
|---|---|---|---|
| WarpBuild | $40 at $0.004/min | $160 at $0.08/min | $200 |
| Blacksmith | $28 after 3,000 free minutes | $160 at $0.08/min | $188 |
| Latchkey | $25 at $0.0025/min | Not offered: $124 on GitHub-hosted | $149 plus $5 to $49 a month for the plan |
The entire $12 difference is Blacksmith's free minutes, which is worth saying plainly: at identical rates, the only thing separating the two on this workload is an allowance. Scale the month up and the gap stays flat at $12 while both bills grow: at 50,000 Linux minutes the Linux lines are $200 against $188, on monthly totals of $360 and $348. Latchkey's own plan is $5 to $49 a month on top of that compute, and its 2,000 to 6,000 included minutes are not netted off, so the figure is gross on both counts.
The two lines that can move it are not on that table. WarpBuild BYOC at $0.002 a minute turns the Linux half into $20 of WarpBuild fees plus whatever your own cloud charges for the instances, which is the cheapest structure on this page for anyone already holding committed spend. And the cache meters run in opposite directions: 50 GB of Blacksmith sticky disk is $25 a month, while 50 GB on WarpBuild is $10 plus about $1 per 10,000 cache operations, though Blacksmith's separate 25 GB weekly actions/cache allowance is free and WarpBuild has no equivalent.
Docker builds: two different products
Both vendors accelerate Docker, and only one of them charges you separately for it. Blacksmith keeps the build on the runner and gives BuildKit a sticky disk per cache key, so the cost is the runner minute plus $0.50 per GB-month of disk. WarpBuild moves the build to a remote builder profile you choose by size, from 16 vCPU at $0.06 a minute upward, which is a second meter running alongside the runner minutes.
Their published claims differ in kind as well as in size. WarpBuild says its builders deliver "up to 50% faster Docker builds compared to traditional solutions"; Blacksmith says "Our customers have reported 2x to 40x improvements in build times". Neither is a benchmark, and the operational details matter more: WarpBuild describes its shared builder cache as eventually consistent with a 10-day TTL, and Blacksmith documents that the first Docker run is uncached and that an idle sticky disk is evicted after 7 days. Our Depot vs Blacksmith for Docker builds page works through the same trade in more detail.
Switching: the exact diff
Both are a label change after one account step, and the account steps differ in an annoying way: Blacksmith installs from the GitHub marketplace, and WarpBuild's bot has to be installed from the WarpBuild dashboard instead. After that, runner selection is per job, so the cheap way to settle this is to point two jobs in the same workflow at the two vendors for a fortnight.
jobs:
test:
- runs-on: ubuntu-latest
+ runs-on: blacksmith-4vcpu-ubuntu-2404
build:
- runs-on: ubuntu-latest
+ runs-on: warp-ubuntu-latest-x64-4x
build-arm:
- runs-on: ubuntu-24.04-arm
+ runs-on: blacksmith-4vcpu-ubuntu-2404-armThe verdict
With identical x64 rates at every size, this comes down to which extras you actually use. Blacksmith wins the default case: 3,000 free minutes a month, arm64 17% cheaper at every size, a 2 vCPU Windows runner WarpBuild does not publish, and a free 25 GB weekly cache allowance against a metered one. WarpBuild wins the specialized cases, and wins them decisively: BYOC at $0.002 a minute inside your own cloud, remote Docker builders you size up to 192 vCPU, and Linux on Apple silicon. Neither publishes a concurrency ceiling worth worrying about, which is a genuine improvement on GitHub-hosted for anyone whose matrices queue. Neither one repairs a job that failed for a reason unrelated to your code, so on both you keep paying for the re-run and the merge keeps waiting for somebody to notice.