# Depot vs WarpBuild for GitHub Actions runners > Depot vs WarpBuild for GitHub Actions runners: the rates stopped matching. Per-minute prices by size, BYOC at $0.002, and a worked monthly bill. Source: https://latchkey.dev/learn/runners/depot-vs-warpbuild Updated: 2026-09-20 Depot vs WarpBuild used to be a tie on price and is not one today: Depot lists a 2 vCPU Linux runner at $0.006 a minute, which is GitHub-hosted list price to the cent, and WarpBuild lists the same shape at $0.004. WarpBuild is a third cheaper at every Linux size both sell, so the case for Depot now has to be made on what the extra money buys. Depot sells ephemeral EC2 runners with a RAM disk in front of the root disk, 2 GB at 2 vCPU rising to 32 GB at 64 vCPU, and a cache its documentation describes at up to 1,000 MiB/s. Usage is tracked by the second with no one-minute minimum, and minutes come bundled into a plan, $20 a month with 2,000 or $200 with 20,000, rather than a standing free tier. WarpBuild sells three things from one account: managed cloud runners at $0.004 a Linux minute, remote Docker builders priced separately from $0.06 a minute, and a bring-your-own-cloud mode at $0.002 a minute where runners execute inside your own AWS, GCP or Azure account and the compute lands on your cloud bill. Billing is per minute, and cache is a line item rather than an allowance. 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. ## Comparison | | Depot | WarpBuild | Latchkey | | --- | --- | --- | --- | | Price per minute, 2 vCPU Linux x64 | $0.006 (2 vCPU, 8 GB RAM, 100 GB disk, 2 GB RAM disk) | $0.004 (2 vCPU, 8 GB RAM) | **$0.0025** (2 vCPU, 8 GB RAM, 100 GB disk) | | Price per minute, 4 vCPU Linux x64 | $0.012 (4 vCPU, 16 GB RAM, 130 GB disk, 4 GB RAM disk) | $0.008 (4 vCPU, 16 GB RAM) | **$0.005** (4 vCPU, 16 GB RAM, 100 GB disk) | | Free tier | None standalone: 2,000 minutes with the $20/month Developer plan, 20,000 with the $200/month Startup plan | Not published. Sign-up needs no card until you reach a usage or credit limit | 30-day trial, then 2,000 included minutes a month on Developer, 4,000 on Launch, 6,000 on Scale | | Billing granularity | Per-second usage tracking, with whole minutes billed at the end of the month and no one-minute minimum | Per minute, in its own words: "CI runners are billed on a per-minute basis" | Per minute | | Runner sizes and OS | 2 to 64 vCPU on Linux x64 and arm64 (Graviton4); Windows Server 2022 and 2025 from 2 vCPU, with no Hyper-V; macOS at 8 vCPU | 2 to 32 vCPU on Linux x64 and arm64; Windows from 4 vCPU; macOS M4 Pro at 6 and 12 vCPU; Linux on Apple silicon 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 | Depot Cache, 25 GB included on Developer and 250 GB on Startup, then $0.20/GB/month; not isolated by branch | Cache storage $0.20 per GB-month plus $0.0001 per write or restore; snapshot restore $0.04 per job and $0.025 per hour of snapshot storage | 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 | A RAM disk on every runner, plus remote container builds as a separate product | Remote Docker builders billed separately, 16 vCPU at $0.06/min up to 192 vCPU at $0.88/min, driven by `Warpbuilds/build-push-action@v7`; cache TTL 10 days | `latchkey-dev/docker-cache-action@v1`, layers held in a Latchkey-managed private registry per organization | | Concurrency limits | Not published | Advertised as unlimited job concurrency, with an explicit no-queue guarantee only on Enterprise | Not published; warm-pool capacity varies by plan | | Where jobs run | Ephemeral EC2 instances in Depot's AWS account, or in your own account on Depot Managed | WarpBuild's own cloud, or your AWS, GCP or Azure account under BYOC at $0.002/min | 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 | Change the `runs-on` label. The repository must be owned by a GitHub organization | Sign up, install the WarpBuild bot from its dashboard rather than the GitHub marketplace, then change the `runs-on` label to a runner ID | A `runs-on` label change. GitHub organization accounts only | ## The rate cards, size by size This is the table the two vendor sites no longer agree on, and it is the reason this page needed rewriting. WarpBuild doubles its rate with every vCPU step from $0.004, and so does Depot from $0.006, so the gap stays at a third all the way up. | Linux x64 | Depot | WarpBuild | GitHub-hosted | Latchkey | | --- | --- | --- | --- | --- | | 2 vCPU | $0.006 | $0.004 | $0.006 | $0.0025 | | 4 vCPU | $0.012 | $0.008 | $0.012 | $0.005 | | 8 vCPU | $0.024 | $0.016 | $0.022 | $0.01 | | 16 vCPU | $0.048 | $0.032 | $0.042 | $0.02 | | 32 vCPU | $0.096 | $0.064 | $0.082 | Not offered | | 2 vCPU arm64 | $0.006 | $0.003 | $0.005 | Not offered | Two things fall out of that table that neither vendor page says out loud. The first is that Depot matches GitHub-hosted list price at 2 and 4 vCPU and then passes it: at 8 vCPU it is $0.024 against GitHub's $0.022, and at 32 vCPU $0.096 against $0.082, because GitHub's ladder flattens as it climbs and Depot's does not. The second is arm64, where WarpBuild charges $0.003 and Depot charges its x64 rate of $0.006, so the architecture that is usually the cheap one costs twice as much on Depot. ## 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 | | --- | --- | --- | --- | | Depot | $60 at $0.006/min | $160 at $0.08/min | $220 before plan fees | | WarpBuild | $40 at $0.004/min | $160 at $0.08/min on a 6 vCPU M4 Pro | $200 | | Latchkey | $25 at $0.0025/min | Not offered: $124 on GitHub-hosted | $149 plus $5 to $49 a month for the plan | The macOS half is identical, which is why the Linux rate decides it: $20 a month on this mix, and $200 a year at three times the volume. Depot also starts from a plan fee of at least $20, where WarpBuild publishes no monthly fee at all, so the gap on a small account is wider than the table shows. 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. BYOC is the line that makes WarpBuild structurally different rather than just cheaper. At $0.002 a minute, the same 10,000 Linux minutes cost $20 in WarpBuild fees, and then your own cloud account pays for the instances, which is a number WarpBuild cannot publish because it is yours. That is the right buy if you hold committed spend or need jobs inside your VPC, and the wrong one if nobody on the team wants another AWS footprint to own. Then the cache meters, which are easy to miss on both. WarpBuild charges $0.20 per GB-month of cache storage plus $0.0001 for every write or restore, so a 50 GB cache is $10 a month plus roughly $1 per 10,000 operations, and a snapshot restore is $0.04 a job on top. Depot includes 25 GB on the Developer plan and charges $0.20 per GB-month after that, so the same 50 GB is $20 of plan plus $5 of storage. Cheap either way at this size, and worth checking before a monorepo pushes it to 500 GB. ## Reliability An Actions SLA appears on these pages only as something you buy: WarpBuild lists a contractual 99.9% uptime SLA and unlimited concurrency under Enterprise, and Depot publishes no figure on its pricing page. Depot does publish one honest constraint worth repeating, which is that Apple licensing makes its macOS capacity inelastic, so macOS jobs can queue at busy times. On both, a job that fails for a reason unrelated to your code fails, bills the minutes it burned, and waits for a person. That is the whole category of work these rate cards do not price, and it is measurable in your own history: count the runs that passed on the second attempt with no code change between them. [Exit code 137](/learn/failures/exit-code-137-in-github-actions) and [npm network errors in CI](/learn/failures/npm-network-errors-in-ci) are the two that most often look like a runner problem and are not. ## When to choose Depot - Your jobs are short and numerous, so per-second tracking with no one-minute minimum beats a lower per-minute rate on a wide matrix. - Your jobs are disk-bound: 100 GB at 2 vCPU with a RAM disk in front of it is more machine than the rate suggests. - You need 64 vCPU Linux, or Windows at 2 vCPU, neither of which WarpBuild publishes. - You want container builds and runner minutes from one vendor, with a cache the docs rate at up to 1,000 MiB/s. ## When to choose WarpBuild - Your bill is mostly Linux minutes. A third off at every size is the plainest saving on this page. - You build on arm64: $0.003 against Depot's $0.006 at 2 vCPU is half price, and below GitHub's own arm64 rate of $0.005. - You want runners inside your own cloud account. BYOC at $0.002 a minute plus your compute is a different cost structure, not a discount. - Your Docker builds want a dedicated remote builder you size yourself, from 16 vCPU at $0.06 a minute up to 192 vCPU at $0.88. - One caveat: Windows starts at 4 vCPU on WarpBuild, and at 4 vCPU the two vendors are level at $0.016 a minute. ## Switching: the exact diff Both are a label change on top of one account step: Depot needs the repository to be owned by a GitHub organization, and WarpBuild needs its bot installed from the WarpBuild dashboard rather than the GitHub marketplace. Nothing else in the workflow moves, and because runner selection is per job you can run this exact diff on two jobs in the same file and compare a fortnight of real runs. ```.github/workflows/ci.yml jobs: test: - runs-on: ubuntu-latest + runs-on: depot-ubuntu-24.04-4 build: - runs-on: ubuntu-latest + runs-on: warp-ubuntu-latest-x64-4x build-arm: - runs-on: ubuntu-24.04-arm + runs-on: warp-ubuntu-latest-arm64-4x ``` ## The verdict On published rates this is now WarpBuild's comparison to lose: a third off Depot at every Linux size, half price on arm64, no plan fee to clear first, and a BYOC mode that changes the cost structure rather than the discount. Depot is not overpriced so much as differently priced, and the two things it sells that WarpBuild does not publish are per-second metering and a genuinely large disk under every runner, both of which matter more than the rate if your matrix is wide or your jobs thrash the filesystem. Price them against your own billed-to-elapsed ratio rather than against each other. Neither of them does anything about the run that failed on a registry timeout at 2am, which is the line item that never appears on a rate card. ## FAQ ### Is WarpBuild cheaper than Depot? On Linux, yes, and by a consistent third: $0.004 against $0.006 at 2 vCPU, $0.008 against $0.012 at 4, and $0.064 against $0.096 at 32, read 20 September 2026. On arm64 it is half: $0.003 against $0.006. macOS is level at $0.08 a minute, and Windows is level at $0.016 for 4 vCPU. Depot can still win on a matrix of very short jobs, because it tracks usage by the second and WarpBuild bills per minute. ### Are Depot runners 3x faster than GitHub-hosted? That is Depot's own claim, published as "up to 3x faster than a GitHub-hosted runner", and it is not a benchmark you can reproduce from the page: it names no workload. What is checkable is the hardware behind it, which is a larger disk, a RAM-disk accelerator and a cache Depot rates at up to 1,000 MiB/s. We have not run our own comparison yet, so we are not repeating a multiplier as though it were measured. ### What is WarpBuild BYOC and what does it actually cost? BYOC runs the runners inside your own AWS, GCP or Azure account while WarpBuild keeps the control plane. The published fee is $0.002 a minute for Linux and for Windows on AWS and Azure, with add-ons free, and the instances themselves appear on your cloud bill at whatever you pay for them. For 10,000 Linux minutes that is $20 in WarpBuild fees plus your own compute, against $40 on its managed runners and $60 on Depot. ### Do Depot and WarpBuild support arm64 runners? Both do, at very different prices. WarpBuild publishes arm64 from $0.003 a minute at 2 vCPU up to $0.048 at 32 vCPU, a 25% discount against its own x64 rate. Depot runs arm64 on Graviton4 and charges the same as x64, $0.006 at 2 vCPU. GitHub-hosted arm64 sits between them at $0.005. Latchkey has no arm64 runner at all, which is a real gap on this page and not one worth papering over. ## References - [Depot GitHub Actions runner types and per-minute prices](https://depot.dev/docs/github-actions/runner-types) - [Depot pricing and plan inclusions](https://depot.dev/pricing) - [WarpBuild pricing: runners, Docker builders, BYOC and add-ons](https://www.warpbuild.com/pricing) - [WarpBuild cloud runner IDs and shapes](https://www.warpbuild.com/docs/ci/cloud-runners) - [GitHub Actions per-minute rates by runner](https://docs.github.com/en/billing/reference/actions-runner-pricing) --- Latchkey runs CI/CD that repairs its own failures. Agent entry points: https://latchkey.dev/agent.txt, https://latchkey.dev/openapi.json, https://latchkey.dev/llms.txt