CircleCI pricing: where is the current price list for credits/minute by compute type (Docker, macOS, Windows)?

CircleCI pricing for credits per minute is published in a single place: the resource-class price tables in the CircleCI docs. Those tables list every compute type (Linux Docker, Remote Docker, Arm, macOS, Windows, GPU, etc.) alongside CPU, RAM, and cost in credits/minute so you can estimate pipeline spend and choose the right executors.

Quick Answer: The current price list for CircleCI credits/minute by compute type lives in the CircleCI documentation under the “Compute” or “Resource class” pricing section. That page is the source of truth for Docker, macOS, Windows, Arm, and GPU credit rates.

The Quick Overview

What It Is: A documented price table that maps each CircleCI resource class (e.g., medium Docker, macos.m4pro.medium, Windows xlarge, GPU machines) to its CPU, RAM, and credits-per-minute cost.
Who It Is For: Engineering leaders, platform teams, and developers who need to control CI/CD spend, model job costs, or standardize on specific resource classes across projects.
Core Problem Solved: It centralizes all compute pricing so you don’t have to guess at credits/minute, scrape UIs, or reverse-engineer costs from invoices.

How It Works

CircleCI uses a credits model: each resource class consumes a fixed number of credits per minute while a job is running. The official price list is maintained in the docs and is updated whenever compute configurations or pricing change.

When you select an executor in your .circleci/config.yml (Docker, machine, macOS, Windows, GPU), you’re really choosing a resource class. That resource class determines:

How many CPUs and how much RAM your job gets.
How many credits per minute the job consumes.
Any additional charges (e.g., Docker Layer Caching, storage, network egress).

From there:

Pick a compute type: Choose Linux Docker, Remote Docker, Arm, macOS, Windows, or GPU based on your app and toolchain.
Select a resource class: Use the docs table to match performance needs (CPU/RAM) to credits/minute for that class.
Estimate and optimize: Multiply credits/minute by typical job duration to estimate cost, then tune jobs to use the smallest reliable class or change architecture (e.g., Arm vs x86).

Credits/minute by compute type (current reference)

Below is a consolidated view based on the current CircleCI resource-class pricing. Always confirm against the live docs for the latest numbers, but this shows how compute types map to credits/minute today.

Linux Docker / Remote Docker (x86)

Standard Linux VM / Remote Docker classes:

Medium: 2 CPU, 7.5 GB RAM — 10 credits/min
Large: 4 CPU, 15 GB RAM — 20 credits/min
X-large: 8 CPU, 32 GB RAM — 100 credits/min
2 X-large: 16 CPU, 64 GB RAM — 200 credits/min
2 X-large+: 32 CPU, 64 GB RAM — 300 credits/min

Gen 2 Linux VM / Remote Docker (x86) classes:

Medium: 2 CPU, 8 GB RAM — 18 credits/min
Large: 4 CPU, 16 GB RAM — 36 credits/min
X-large: 8 CPU, 32 GB RAM — 72 credits/min
2 X-large: 16 CPU, 64 GB RAM — 144 credits/min
2 X-large+: 32 CPU, 128 GB RAM — 288 credits/min

Docker / Remote Docker (Arm)

Arm-based Docker and Remote Docker classes:

Medium: 2 CPU, 8 GB RAM — 13 credits/min
Large: 4 CPU, 16 GB RAM — 26 credits/min
X-large: 8 CPU, 32 GB RAM — 52 credits/min
2 X-large: 16 CPU, 64 GB RAM — 104 credits/min

Linux VM (Arm-only)

Arm VM classes:

Medium: 2 CPU, 8 GB RAM — 10 credits/min
Large: 4 CPU, 16 GB RAM — 20 credits/min
X-large: 8 CPU, 32 GB RAM — 100 credits/min
2 X-large: 16 CPU, 64 GB RAM — 200 credits/min

macOS VM

macOS M4 Pro classes:

M4 Pro Medium: 6 CPU, 28 GB RAM — 200 credits/min
M4 Pro Large: 12 CPU, 56 GB RAM — 400 credits/min

These are typical for iOS/macOS builds and UI tests.

Windows VM

Windows VM classes:

Medium: 4 CPU, 16 GB RAM — 40 credits/min
Large: 8 CPU, 32 GB RAM — 120 credits/min
X-large: 16 CPU, 64 GB RAM — 210 credits/min

GPU compute

GPU-accelerated resource classes for AI/ML or GPU-heavy workloads:

Medium Nvidia Tesla T4 GPU
- 8 CPU, 30 GB RAM — 240 credits/min
Medium Multi 4 Nvidia Tesla T4 GPUs
- 8 CPU, 30 GB RAM — 240 credits/min (multi-GPU configuration)
Windows Medium Nvidia Tesla T4 GPU
- 16 CPU, 60 GB RAM — 500 credits/min
Large Nvidia Tesla V100 GPU
- 8 CPU, 30 GB RAM — 1,000 credits/min

Non-compute pricing (related to job cost)

These items are billed in addition to the base compute credits/min when used:

Docker Layer Caching (DLC): 200 credits/job
IP ranges: 450 credits/GB
Runner network egress: 420 credits/GB
Storage: 420 credits/GB

If you’re modelling end-to-end pipeline cost, factor these alongside resource-class usage.

Features & Benefits Breakdown

Core Feature	What It Does	Primary Benefit
Centralized compute pricing	Publishes all credits/minute rates by resource class in one doc table	Clear cost visibility across Docker, macOS, Windows, GPU
Resource-class granularity	Shows CPU/RAM per class so you can match workloads to the right executor	Avoid overprovisioning while keeping builds and tests stable
Multi-architecture options	Lists x86, Arm, macOS, Windows, and GPU pricing side by side	Lets platform teams standardize golden paths per stack

Ideal Use Cases

Best for delivery teams planning budgets: Because it lets you estimate pipeline credits by workload type (Linux, macOS, Windows, GPU) and pick resource classes that keep costs in line with SLAs.
Best for platform teams defining golden paths: Because it gives you a concrete menu of approved resource classes and their prices to codify in templates, contexts, and policy checks.

Limitations & Considerations

Pricing can change: The numbers above reflect the current docs snapshot; you should always verify against the live CircleCI pricing/resource-class page before locking in budgets.
Regional or feature nuances: Some compute types or add-ons (like DLC or specific GPU SKUs) may have availability constraints or require configuration changes, so validate that they’re enabled in your org before standardizing on them.

Pricing & Plans

CircleCI plans (Free, Performance, Scale, enterprise agreements) determine how many credits you get and at what commercial rate, but the credits-per-minute per resource class is the same schedule you see in the compute pricing tables.

Most teams work like this:

Use the credits/minute tables to model per-job and per-workflow cost.
Pick a CircleCI plan that gives enough credits for their expected monthly usage.
Iterate on pipeline design (Smarter Testing, Chunk, right-sized executors, rollback pipelines) to increase throughput per credit.

Details on plan pricing and credit bundles are available on the main CircleCI pricing page, while the resource-class tables remain the authoritative list for credits/minute.

Performance-style plans: Best for product teams and scale-ups needing predictable credits, varied compute types (Docker, macOS, Windows), and fast feedback loops.
Scale/Enterprise-style plans: Best for larger organizations needing negotiated credit volumes, governance tooling (Platform Toolkit, policy checks), and standardized resource-class usage across many repos.

Frequently Asked Questions

Where is the official CircleCI credits-per-minute price list?

Short Answer: In the CircleCI documentation under the compute/resource-class pricing section.

Details: CircleCI maintains a single, canonical set of tables in the docs that list each resource class by compute type (Linux Docker, Remote Docker, Arm, macOS, Windows, GPU) with CPU, RAM, and credits/minute. That page is kept up to date as new executors are added or pricing changes. For financial planning, compliance reviews, or internal documentation, always reference that doc rather than relying on cached UIs or old screenshots.

How do I estimate the credit cost of a CircleCI job?

Short Answer: Multiply the resource class’s credits/minute by the job’s runtime, then add any add-on costs like DLC or storage.

Details: Suppose your main build runs on a Linux Docker large class at 20 credits/min and averages 5 minutes. That job consumes about 100 credits per run. If you enable Docker Layer Caching, add 200 credits per job. For workflows with multiple jobs, repeat the calculation per job (credits/min × average duration) and sum them. This lets you compare alternatives—for example, a shorter job on a higher class vs a longer job on a smaller class—and choose the path that delivers trusted signal at acceptable cost.

Summary

The current price list for CircleCI credits per minute by compute type is exposed through the resource-class tables in the CircleCI docs. Those tables are your source of truth for mapping Docker, macOS, Windows, Arm, and GPU executors to their CPU, RAM, and credit rates. Once you have that mapping, you can design golden-path pipelines that balance AI-speed delivery with spend control: right-sized resource classes, fast tests, safe rollbacks, and governance via policy checks and approvals.

Next Step

Get Started