M5 Max runners are here: macOS and ARM Linux CI on Apple Silicon

Avrea now supports GitHub Actions workloads on macOS and ARM Linux powered by the latest M5 Max Apple Silicon. That means native Apple builds, fast ARM Linux jobs and access to some of the fastest ARM CPUs available for CI with built-in caching to make it even faster.

macOS and ARM CI should not be the slow lane

For most teams, these are the workloads CI does worst. You either wait for constrained hosted macOS capacity, or run ARM workloads on infrastructure that was not really built for fast CI. Avrea's M5 Max runners change that.

For comparison, as of writing this, GitHub offers macOS M2 Pro (5-core) runners as the fastest machine and as a default macOS option an M1 3-core. GitHub charges $0.102 per minute for M2 Pro 5-core where Avrea charges $0.08 for the M5 Max 8-core.

Why we chose M5 Max

M5 Max gives CI workloads something they usually do not get: high-end Apple Silicon CPUs as infrastructure, not just as developer laptops.

For many build-heavy jobs, CPU performance still matters more than almost anything else. Compilers, package managers, test runners, bundlers, linkers, and compression tools all spend meaningful time on CPU-bound work.

M5 Max is especially interesting because it brings some of the fastest ARM CPU cores available, high single-thread performance for build steps that do not parallelize well, strong multi-core throughput for compilation and test suites, unified memory bandwidth for large build graphs and dependency-heavy workloads, and native execution for macOS, iOS, Swift, Xcode, and Apple-platform builds.

Available macOS and ARM Linux runner sizes

M5 Max macOS runners are available in two sizes:

M5 Max ARM Linux runners are available in five sizes:

The labels make CPU and memory explicit in the workflow. Start with the size that matches your build, monitor CPU and RAM usage in Avrea, and scale up when the job is actually bottlenecked. Larger labels are useful for compilation, test suites, packaging, and multi-architecture release jobs that can take advantage of more CPU or memory.

What this makes faster

Fast ARM CPUs help, but CI performance is not only about raw CPU time. A lot of build time is spent downloading dependencies, restoring package manager state, rebuilding unchanged artifacts, and moving cache data over the network.

That is where Avrea's caching support matters.

For macOS builds, in addition to the many caching types Avrea supports, we also support Xcode caching so repeated macOS, iOS, watchOS, visionOS, and Swift workflows can avoid rebuilding work that has already been done. This is especially useful for dependency-heavy Xcode projects where derived data, compiled modules, package state, and generated build outputs can dominate repeated CI runs.

For ARM Linux workloads, Avrea supports caching across the ecosystems teams commonly use in CI. Go builds can benefit from cached module and build outputs. Rust builds can reuse Cargo registry data, git dependencies, and compiled artifacts. Node.js and TypeScript projects can avoid repeatedly downloading npm, pnpm, or Yarn dependencies. Python projects can reuse package and build caches. C, C++, and Zig projects can benefit from compiler and build-system caches where the workflow is configured to use them.

Avrea also supports GitHub Actions cache, build cache, and package cache, so teams can accelerate the parts of the workflow around the compiler too: dependency installation, package downloads, generated artifacts, and repeated build steps across branches and pull requests.

The result is that M5 Max runners are not just faster machines, but faster machines with caching designed to reduce the repeated work that makes macOS and ARM Linux CI slow.

Use the same workflow on faster runners

Avrea runners work with GitHub Actions workflows, so teams can usually start by changing the runner label instead of redesigning their pipeline.

jobs:
 build:
   runs-on: avrea-macos-26-16-vcpu
   steps:
     - uses: actions/checkout@v6
     - run: xcodebuild test


or:

jobs:
 arm64-build:
   runs-on: avrea-ubuntu-latest-arm-16-vcpu
   steps:
     - uses: actions/checkout@v6
     - run: make test


Migration only takes 5 minutes. Keep the workflow structure, update the label, and run the job on Avrea's M5 Max infrastructure.

Sign up for Avrea and try M5 Max macOS and ARM Linux runners in your GitHub Actions workflows.