# Fast GitHub Actions Runners Build 2x faster than on GitHub-hosted runners on hardware alone, and 10x faster with cache hits. Run CI on Linux, macOS and Windows. GitHub Actions Runners Build 2x faster than on GitHub-hosted runners on hardware alone, and 10x faster with cache hits. Run CI on Linux, macOS and Windows, without managing a runner fleet. [Start free](https://console.avrea.com/) [See pricing](https://avrea.com/pricing) *Linux kernel build, ccache* GitHub Runner · 2 vCPU 27m 23s Avrea Runner · 2 vCPU 9m 3s Avrea Runner · 2 vCPU, cached 24.1s 3.0x faster 68x faster, cached ## Purpose-built hardware for CI runners Dedicated hardware and Avrea’s runner technology deliver 2x faster builds than GitHub-hosted runners on compute alone. - ### Fast CPUs AMD EPYC 4585PX processors for Linux x86 runners and Apple M5 Max for macOS and Linux ARM. - Avrea5,800+ - Cloud VM~3,200 - GitHub~2,500 PassMark single-thread - ### NVMe storage Local NVMe storage for fast cache restores and artifact access. - Avrea~7 GB/s - Cloud VM~2 GB/s - GitHub~0.7 GB/s Sequential read - ### 10 Gbps+ network Fast transfers for dependencies, containers and build artifacts. - Avrea10 Gbps+ - Cloud VM~2 Gbps - GitHub~1 Gbps Sustained throughput - ### Dedicated bare metal Each job runs in an isolated VM on Avrea’s dedicated hardware. Isolated job VM Avrea hardware ### Caching at every layer Builds run 10x faster than on GitHub-hosted runners with cache hits. Avrea caches Git data, packages, Docker layers and build outputs alongside your runners. Package cache 7s from 16s npm, Go, pip, cargo Docker layer cache 4.5s from 111s BuildKit type=gha GitHub Actions cache <5s from 15s actions/setup-node Build cache 13s from 36s Go, Rust, C++, Bazel, Turbo Hit rate94% Storage8 / 25 GB Artifacts12,483 ### Runners for every platform Build and test on Linux, macOS and Windows, with native x86 and ARM runners. - Linux x86 · ARM - macOS Apple Silicon - Windows x86 ## Built for scale Run CI across your organization with high concurrency, private network access and a fully managed runner fleet. ### High concurrency Run large build matrices and test suites in parallel across your repositories. ### Connect to your private network Reach private registries, databases and internal services through your VPN. Set outbound access rules for each organization or repository. ### The benefits of self-hosting, without the maintenance Get dedicated hardware and persistent caches while Avrea handles provisioning, scaling and maintenance. Every job runs in an isolated VM. Avrea CLI $avr workflow run ci.yml $avr run list --json status ### CI from your terminal Trigger workflows, rerun failed jobs and retrieve logs through the Avrea CLI, with structured output for scripts and agents. ## Full CI observability *Job metrics · CPU, memory, disk and network* ### Build times and reliability Track build times, failures and flaky jobs across your repositories. ### Runner right-sizing Avrea checks what each job uses and recommends the right runner size. ### Searchable logs and usage Search step logs, follow live output and track runner usage by workflow. ### Debug a running job SSH into a running job to inspect its environment and investigate failures. [Explore Observability](https://avrea.com/observability) ## Switch runners with a label change Connect Avrea to GitHub and update your workflow’s `runs-on` label. Keep your existing workflows, actions and secrets. [Read the setup guide](https://docs.avrea.com/getting-started/) [See runner pricing](https://avrea.com/pricing) GitHub Actions · runner label jobs: test: - runs-on: ubuntu-latest + runs-on: avrea-ubuntu-latest-2-vcpu ## Lower your CI costs Lower per-minute pricing for Linux x86 runners. Faster builds and cache hits can reduce your billed runtime further. [See runner pricing](https://avrea.com/pricing#runner-pricing) *33% lower per minute* 2 vCPU · Linux x86 · USD Avrea $0.004 / min GitHub $0.006 / min Published list rates · [GitHub pricing](https://docs.github.com/en/billing/reference/actions-runner-pricing) Benchmarks ## Faster builds on Avrea - ### Linux kernel Benchmark study Avrea · cached 24.1s Avrea · cold 9m 3s GitHub Actions 27m 23s Without cache 3.0x faster than GitHub Actions With cache 68x faster than GitHub Actions - ### Bazel Benchmark study Avrea · cached 56.1s Avrea · cold 14m 4s GitHub Actions 37m 10s Without cache 2.6x faster than GitHub Actions With cache 39.7x faster than GitHub Actions - ### Ghostty Benchmark study Avrea · cached 23.0s Avrea · cold 3m 33s GitHub Actions 10m 27s Without cache 2.9x faster than GitHub Actions With cache 27x faster than GitHub Actions - ### Next.js Benchmark study Avrea · cached 1.4s Avrea · cold 1m 23s GitHub Actions 3m 23s Without cache 2.4x faster than GitHub Actions With cache 142x faster than GitHub Actions - ### RisingWave Benchmark study Avrea · cached 1m 52s Avrea · cold 10m 48s GitHub Actions 21m 40s Without cache 2.0x faster than GitHub Actions With cache 11.5x faster than GitHub Actions - ### Kafka Benchmark study Avrea · cached 52.2s Avrea · cold 2m 10s GitHub Actions 5m 44s Without cache 2.6x faster than GitHub Actions With cache 6.6x faster than GitHub Actions View benchmark times as a table Build duration by workload and cache state. Each workload links to its benchmark study. *Benchmark build durations* | Workload | GitHub Actions | Avrea cold | Avrea cached | | --- | --- | --- | --- | | [Linux kernel](https://avrea.com/blog/linux-kernel-ci-benchmark) | 27m 23s | 9m 3s | 24.1s | | [Bazel](https://avrea.com/blog/bazel-ci-benchmark) | 37m 10s | 14m 4s | 56.1s | | [Ghostty](https://avrea.com/blog/ghostty-nix-ci-benchmark) | 10m 27s | 3m 33s | 23.0s | | [Next.js](https://avrea.com/blog/nextjs-turborepo-ci-benchmark) | 3m 23s | 1m 23s | 1.4s | | [RisingWave](https://avrea.com/blog/risingwave-sccache-ci-benchmark) | 21m 40s | 10m 48s | 1m 52s | | [Kafka](https://avrea.com/blog/kafka-gradle-ci-benchmark) | 5m 44s | 2m 10s | 52.2s | Results are specific to each workload, runner configuration and cache state. ## Run your GitHub Actions jobs on Avrea [Start free](https://console.avrea.com/) [Talk to an engineer](https://avrea.com/demo)