CircleCI pricing 2026
./circleci --credits-explained
CircleCI is the only major CI/CD platform that bills compute in credits instead of minutes. The system is more flexible than per-minute pricing because larger machines burn credits faster, but it's also opaque if you've never thought about resource classes. This guide demystifies the credits model and shows what CircleCI actually costs at every team size.
How much does CircleCI cost in 2026?
The Free plan gives you 30,000 credits a month (about 3,000 Linux Medium build minutes) with up to 5 users. The paid Performance plan starts at $15/month, including 30,000 credits plus 5 active users; each extra user is $15 and extra credits cost $15 per 25,000 pack, roughly $0.0006 per credit. Because CircleCI bills in credits, your real cost depends on the resource class: Linux Medium burns 10 credits per minute ($0.006), while macOS M4 Pro runs 200 to 400 credits per minute ($0.12 to $0.24). Scale is a custom annual contract for enterprises.
Credit cost by resource class
| Resource class | vCPU / RAM | Credits / min | Cost / min | 10K mins cost |
|---|---|---|---|---|
| Small (Linux) | 1 / 2 GB | 5 | $0.003 | $30 |
| Medium (Linux) | 2 / 4 GB | 10 | $0.006 | $60 |
| Large (Linux) | 4 / 8 GB | 20 | $0.012 | $120 |
| X-Large (Linux VM) | 8 / 32 GB | 100 | $0.060 | $600 |
| Medium (Windows) | 4 / 16 GB | 40 | $0.024 | $240 |
| M4 Pro Medium (macOS) | 6 / 28 GB | 200 | $0.120 | $1,200 |
| M4 Pro Large (macOS) | 12 / 56 GB | 400 | $0.240 | $2,400 |
Plans / 2026
Worked example / 20 dev team
The included 30,000 credits are an org-level pool, not per user, so a mid-size team burns through them in the first few days of the month. Most of the bill is active-user charges plus credit packs. Push to large or x-large machines, or switch resource classes to macOS, and the credit packs stack up fast.
How to lower CircleCI spend
- > Right-size resource_class in .circleci/config.yml. Defaults to medium; small often works for unit tests.
- > Use parallelism: with split tests to cut wall-clock time without raising machine size.
- > Cache aggressively with restore_cache and save_cache for dependency installs.
- > Self-host runners for macOS workloads or sustained heavy compute.
- > Watch the Insights dashboard for the longest jobs, optimise those first.
Compare with other platforms
Frequently Asked Questions
# click any question to expand