$ ~/cicdcalculator

CI/CD cost at 25,000 build minutes per month, 2026
./tier --monthly-mins=25000 --25-dev-team

Twenty-five thousand monthly Linux build minutes is the typical CI consumption of a 20-25 developer engineering team running 6-8 minute builds on every pull request push. It is the volume at which the differences between CI/CD pricing models become most visible: per-minute platforms start to bill noticeable overage, self-hosted starts to look financially attractive, and the seat-fee-plus-allowance model on platforms like GitLab Premium begins competing with the per-minute model on Actions and CircleCI. This page walks through what each major platform charges at exactly 25,000 monthly minutes in 2026.

All numbers below assume Linux 2-core hosted runners, a 25-person team where seat fees apply, and 25,000 monthly minutes total. Source pricing pulled from each vendor's pricing page in May 2026.

Side-by-side at 25,000 minutes

PlatformPlanSeats (25 users)ComputeTotal / mo
GitHub ActionsTeam$100$220 (22K over 3K free)$320
GitLab CIPremium$725$150 (15K over 10K)$875
CircleCIPerformance$375$0 (in 625K credits)$375
Bitbucket PipelinesPremium$150$190 (19K over 6K)$340
BuildkitePro$375$180 (4 EC2 agents)$555
JenkinsSelf-hosted$0$200 (controller + agents)$200 + admin
TeamCityServer (8 agents)$0$340 / mo amortised$340 + infra + admin

# Self-hosted totals exclude operator time (4-8 hrs/mo at this scale = $600-1,200/mo implicit cost). Bitbucket assumes Premium for higher included allowance.

The cheapest options at this tier

Self-hosted Jenkins or Drone CI on AWS at around $200 per month is the cheapest direct-cost answer. Add the implicit operator cost and the picture becomes more nuanced; at this volume, hosted platforms are competitive once full-cost accounting is honest.

Among hosted options, Bitbucket Pipelines Premium at $340 per month and GitHub Actions Team at $320 per month are both competitive. Bitbucket is slightly cheaper on raw seats; GitHub Actions slightly cheaper on minute overage. The difference is small enough that other factors (source-control choice, ecosystem) dominate the decision.

CircleCI Performance at $375 per month is competitive because the included 625,000 credits per team easily absorbs the 250,000-credit equivalent of 25,000 Linux Medium minutes. CircleCI starts to look worse at higher build volumes when overage credits eat into the per-user allowance, but at 25K minutes the value is solid.

Where the seat-fee model loses

GitLab Premium at $725 per month for 25 seats is hard to justify at 25,000 monthly minutes unless you specifically need the Premium feature set (advanced merge approvals, value-stream analytics, deploy environments). The seat premium versus GitHub Actions Team is roughly 7x at this scale, which is real money. Most teams that pick GitLab Premium do so because they want the features, not because the CI/CD pricing is competitive.

Buildkite Pro at $555 per month is similarly more expensive than per-minute hosted alternatives at this volume. Buildkite's value proposition is volume-dependent: above 50,000 monthly minutes the per-seat-plus-BYO model wins, below that it does not. At exactly 25K minutes, Buildkite is competitive only for teams who specifically benefit from its strengths (mobile, data-residency, very high parallelism on stable agent pools).

Decision rules at 25,000 minutes

Default to GitHub Actions Team if you are on GitHub for source. The pricing is competitive, the ecosystem is enormous, and the operational simplicity at this scale is hard to beat. Reach for self-hosted runners (free at the platform level) when build volume keeps climbing past this tier.

Default to GitLab CI Free with self-hosted runners if you are on GitLab. The Premium plan only makes sense if you actually need Premium features. Premium for the CI/CD allowance alone is overpriced at this scale.

Consider self-hosted Jenkins or Drone if you have an existing platform-engineering function and want to consolidate CI/CD into infrastructure you already operate. Below this scale the operator cost dominates; above this scale self-hosted increasingly wins.

Skip Buildkite at this exact tier unless you ship iOS or have data-residency needs. Buildkite is a strong long-term commitment but does not earn its premium at 25K minutes for typical Linux-only workloads.

Optimisation before upgrading

Before moving to a more expensive tier, optimise. Aggressive dependency caching typically cuts billable minutes by 30-50 percent. Conditional jobs (skip frontend tests on backend-only PRs) often add another 10-20 percent. Right-sized runners (a 4-core finishing in 4 minutes can be cheaper than a 2-core in 12 minutes) help further. Many teams that think they have a 25,000-minute workload actually have a 12,000-minute workload after a careful optimisation pass.

See the cost optimization deep dive for the ten techniques that move the needle most. Most are platform-agnostic and apply equally to GitHub Actions, GitLab CI, CircleCI and self-hosted Jenkins.

Other tiers and platform deep dives

Frequently Asked Questions

# click any question to expand

Why is 25,000 monthly minutes the right benchmark?>
It is the build volume of a typical 20-25 person engineering team running a moderately complex pipeline (4-6 stages, 8 minute average) on every push. The number falls just outside the GitHub Actions Team plan free allowance and inside the GitLab Premium allowance, which makes the cost differences between platforms most visible. Smaller teams sit below this; larger teams or test-heavy teams sit above.
Which platform wins at 25K minutes?>
Self-hosted runners on any platform win on raw cost ($150-200 per month for the infrastructure plus $300-600 in operator time). Among hosted-only options, Bitbucket Pipelines Premium is competitive at around $340 per month for a 25-dev team; GitHub Actions Team comes in around $370. CircleCI Performance and GitLab Premium are higher because of richer included credit allowances vs higher seat fees.
When does Buildkite become competitive?>
Just above this volume. At 25,000 minutes Buildkite Pro for a 25-dev team is around $555 ($375 in seats plus $180 in agent compute). That is more than the per-minute platforms at this scale. Push past 50,000 minutes and Buildkite catches up because seat fees stay flat while per-minute platforms compound. iOS-shipping teams flip the math sooner because hosted macOS is so expensive everywhere.
Is self-hosted worth it at 25K minutes?>
Marginal. The dollar saving is real (around $200-300 per month versus the cheapest hosted option) but the operator time required (4-8 hours monthly) often eats the saving when costed at full engineering rates. Self-hosted makes more obvious sense at higher volume, or for teams where the operator time is absorbed into existing platform-engineering capacity rather than being a marginal hire decision.
What about hybrid setups?>
Hybrid (hosted for PR checks plus self-hosted for big pipelines) starts to make sense around this volume. The hosted side handles elastic peak load and short jobs; the self-hosted side handles the bulky integration tests and release builds. Hybrid is operationally more complex than either pure model but absorbs the strengths of both.
How does macOS in the mix change the math?>
Dramatically. If 5,000 of the 25,000 monthly minutes are macOS, hosted-platform costs roughly double. GitHub Actions macOS at $0.082 per minute on top of platform fee adds $410 to the bill, on top of the Linux baseline. Self-hosted Mac minis on Buildkite or GitHub Actions add roughly $50-75 per Mac mini per month and absorb the same workload. For mobile teams running serious build volume, self-hosted Mac is the answer almost regardless of other considerations.
Can I optimise the build to drop below 25K minutes?>
Often yes. The biggest wins are dependency caching (npm, gradle, pip caches typically halve install time), conditional jobs (skip frontend tests on backend-only PRs), test parallelism with proper rebalancing, and right-sizing runners (a 4-core machine that finishes in 4 minutes can be cheaper than a 2-core in 12). Most teams that optimise carefully cut billable minutes by 30-50 percent. See the cost optimization deep dive for ten proven techniques.