RunsOn runner benchmark.
4620 of 4729 jobs finished across every suite, 109 failed (Rust 3, TypeScript 26, Docker 3, Hardware 10, Burst 65, Cache 2).
- TypeScript is borderline at 4 GB or less: RunsOn's c8a.large · EBS gp3, c8g.large · EBS gp3 and t8i.medium · EBS gp3 ran out of memory; boat never ran out; AWS CodeBuild ran out too; 8 GB passes on RunsOn.
- Measured swap at 4 GB or less: boat 2 vCPU has 6 GiB and RunsOn 2 vCPU 3.7 GiB; AWS CodeBuild 2 vCPU has none.
Fastest variant per build, any size
- Fastest x64 Rust build: RunsOn m8azn.3xlarge · EBS gp3 (12 vCPU) in 57s, ahead of Avrea 8 vCPU (8 vCPU, 1m 04s).
Fastest variant per build, any size
Suggested default at 4 vCPU
12m 00s and $0.0218 at spot for the three builds: 5% slower than m8azn.xlarge, the fastest, and 46% cheaper.
How it's picked
Among RunsOn's 4 vCPU variants on the default EBS volume that finished all three builds: the cheapest for all three at the selected price (spot or on-demand), within 10% of the fastest one's time.Suggested default at 4 vCPU
12m 00s and $0.0626 at on-demand for the three builds: 5% slower than m8azn.xlarge, the fastest, and 36% cheaper.
How it's picked
Among RunsOn's 4 vCPU variants on the default EBS volume that finished all three builds: the cheapest for all three at the selected price (spot or on-demand), within 10% of the fastest one's time.Suggested default at 4 vCPU
m9g.xlarge · EBS gp3 arm64
11m 51s and $0.0231 at spot for the three builds: the only variant on the default EBS volume that finished all three.
How it's picked
Among RunsOn's 4 vCPU variants on the default EBS volume that finished all three builds: the cheapest for all three at the selected price (spot or on-demand), within 10% of the fastest one's time.Suggested default at 4 vCPU
m9g.xlarge · EBS gp3 arm64
11m 51s and $0.0508 at on-demand for the three builds: the only variant on the default EBS volume that finished all three.
How it's picked
Among RunsOn's 4 vCPU variants on the default EBS volume that finished all three builds: the cheapest for all three at the selected price (spot or on-demand), within 10% of the fastest one's time.Variants
Fastest firstCheapest first; × against RunsOn's fastest variant with 3 runs; cost per build at EC2 spoton-demand, on the whole job's time.
+19 more variants
+19 more variants
+19 more variants
+3 more variants
+3 more variants
+3 more variants
Bar shades, light to dark, in execution order: setup = checkout + packages + toolchaincargo fmtcargo testclippy
Bar shades, light to dark, in execution order: toolchainclonebun installtypecheck
Bar shades, light to dark, in execution order: fetch sourcebuildx setupdocker build
Same CPU, different disk
- m8a.2xlarge · EBS gp3 and m8a.2xlarge · gp3 1000 MiB/s m8a.2xlarge on EBS gp3 and on gp3 1000 MiB/s: Rust even within run-to-run spread (1m 13s–1m 21s on EBS gp3, 1m 12s–1m 21s on gp3 1000 MiB/s); TypeScript 35s on gp3 1000 MiB/s vs 45s; Docker even within run-to-run spread (5m 33s–6m 28s on EBS gp3, 4m 13s–7m 11s on gp3 1000 MiB/s).
- m8a.2xlarge · EBS gp3 and m8a.2xlarge · tmpfs m8a.2xlarge on EBS gp3 and on tmpfs: Rust even within run-to-run spread (1m 13s–1m 21s on EBS gp3, 1m 13s–1m 18s on tmpfs); TypeScript 32s on tmpfs vs 45s; Docker 3m 32s on tmpfs vs 5m 39s.
- m8a.2xlarge · gp3 1000 MiB/s and m8a.2xlarge · tmpfs m8a.2xlarge on gp3 1000 MiB/s and on tmpfs: Rust even within run-to-run spread (1m 12s–1m 21s on gp3 1000 MiB/s, 1m 13s–1m 18s on tmpfs); TypeScript even within run-to-run spread (34s–35s on gp3 1000 MiB/s, 29s–36s on tmpfs); Docker 3m 32s on tmpfs vs 4m 39s.
- m8azn.xlarge · gp3 1000 MiB/s and m8azn.xlarge · EBS gp3 m8azn.xlarge on gp3 1000 MiB/s and on EBS gp3: Rust even within run-to-run spread (1m 37s–1m 48s on gp3 1000 MiB/s, 1m 36s–1m 47s on EBS gp3); TypeScript 47s on gp3 1000 MiB/s vs 57s; Docker even within run-to-run spread (6m 38s–9m 53s on gp3 1000 MiB/s, 7m 50s–21m 49s on EBS gp3).
- r8a.large · EBS gp3 and r8a.large · tmpfs r8a.large on EBS gp3 and on tmpfs: Rust even within run-to-run spread (2m 50s–3m 15s on EBS gp3, 2m 53s–3m 20s on tmpfs); TypeScript even within run-to-run spread (1m 35s–1m 47s on EBS gp3, 1m 25s–1m 38s on tmpfs).
- m8i.2xlarge · EBS gp3 and m8id.2xlarge · local NVMe m8i.2xlarge (EBS gp3) and m8id.2xlarge (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (1m 38s–1m 55s on EBS gp3, 1m 40s–1m 53s on local NVMe); TypeScript 45s on local NVMe vs 57s; Docker even within run-to-run spread (6m 50s–9m 00s on EBS gp3, 5m 42s–8m 32s on local NVMe).
- m8i-flex.2xlarge · EBS gp3 and m8id.2xlarge · local NVMe m8i-flex.2xlarge (EBS gp3) and m8id.2xlarge (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (1m 41s–2m 02s on EBS gp3, 1m 40s–1m 53s on local NVMe); TypeScript 45s on local NVMe vs 1m 02s; Docker even within run-to-run spread (7m 21s–9m 33s on EBS gp3, 5m 42s–8m 32s on local NVMe).
- m8i-flex.large · EBS gp3 and m8id.large · local NVMe m8i-flex.large (EBS gp3) and m8id.large (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (4m 39s–5m 41s on EBS gp3, 4m 38s–5m 09s on local NVMe); TypeScript even within run-to-run spread (2m 48s–3m 25s on EBS gp3, 2m 46s–3m 34s on local NVMe).
- m8i-flex.xlarge · EBS gp3 and m8id.xlarge · local NVMe m8i-flex.xlarge (EBS gp3) and m8id.xlarge (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (2m 40s–2m 57s on EBS gp3, 2m 37s–2m 49s on local NVMe); TypeScript 1m 20s on local NVMe vs 1m 31s; Docker 9m 15s on local NVMe vs 11m 46s.
- m8i.large · EBS gp3 and m8id.large · local NVMe m8i.large (EBS gp3) and m8id.large (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (4m 38s–5m 37s on EBS gp3, 4m 38s–5m 09s on local NVMe); TypeScript even within run-to-run spread (2m 32s–3m 17s on EBS gp3, 2m 46s–3m 34s on local NVMe).
- m8i.xlarge · EBS gp3 and m8id.xlarge · local NVMe m8i.xlarge (EBS gp3) and m8id.xlarge (local NVMe) share the Xeon 6975P-C: Rust even within run-to-run spread (2m 41s–2m 51s on EBS gp3, 2m 37s–2m 49s on local NVMe); TypeScript even within run-to-run spread (1m 24s–1m 34s on EBS gp3, 1m 19s–1m 24s on local NVMe); Docker 9m 15s on local NVMe vs 11m 56s.
- m9g.2xlarge · tmpfs and m9g.2xlarge · EBS gp3 m9g.2xlarge on tmpfs and on EBS gp3: Rust even within run-to-run spread (1m 13s–1m 20s on tmpfs, 1m 13s–1m 23s on EBS gp3); TypeScript 31s on tmpfs vs 46s; Docker even within run-to-run spread (3m 57s–7m 21s on tmpfs, 5m 44s–6m 13s on EBS gp3).
- m9g.xlarge · EBS gp3 and m9g.xlarge · tmpfs m9g.xlarge on EBS gp3 and on tmpfs: Rust even within run-to-run spread (1m 47s–1m 52s on EBS gp3, 1m 46s–2m 14s on tmpfs); TypeScript 48s on tmpfs vs 58s.
- m9g.2xlarge · tmpfs and m9gd.2xlarge · local NVMe m9g.2xlarge (tmpfs) and m9gd.2xlarge (local NVMe) share the Neoverse-V3: Rust even within run-to-run spread (1m 13s–1m 20s on tmpfs, 1m 11s–1m 17s on local NVMe); TypeScript even within run-to-run spread (28s–38s on tmpfs, 30s–34s on local NVMe); Docker even within run-to-run spread (3m 57s–7m 21s on tmpfs, 4m 17s–5m 12s on local NVMe).
- m9g.xlarge · EBS gp3 and m9gd.xlarge · local NVMe m9g.xlarge (EBS gp3) and m9gd.xlarge (local NVMe) share the Neoverse-V3: Rust even within run-to-run spread (1m 47s–1m 52s on EBS gp3, 1m 45s–1m 56s on local NVMe); TypeScript 50s on local NVMe vs 58s; Docker 6m 19s on local NVMe vs 9m 04s.
- m9g.2xlarge · EBS gp3 and m9gd.2xlarge · local NVMe m9g.2xlarge (EBS gp3) and m9gd.2xlarge (local NVMe) share the Neoverse-V3: Rust even within run-to-run spread (1m 13s–1m 23s on EBS gp3, 1m 11s–1m 17s on local NVMe); TypeScript 30s on local NVMe vs 46s; Docker 4m 24s on local NVMe vs 6m 03s.
- m9g.large · EBS gp3 and m9gd.large · local NVMe m9g.large (EBS gp3) and m9gd.large (local NVMe) share the Neoverse-V3: Rust even within run-to-run spread (2m 50s–3m 07s on EBS gp3, 2m 50s–2m 56s on local NVMe); TypeScript even within run-to-run spread (1m 39s–2m 01s on EBS gp3, 1m 40s–1m 53s on local NVMe).
- m9g.xlarge · tmpfs and m9gd.xlarge · local NVMe m9g.xlarge (tmpfs) and m9gd.xlarge (local NVMe) share the Neoverse-V3: Rust even within run-to-run spread (1m 46s–2m 14s on tmpfs, 1m 45s–1m 56s on local NVMe); TypeScript even within run-to-run spread (46s–53s on tmpfs, 49s–51s on local NVMe).
Also differs in RAM (13)
- c8a.large · EBS gp3 and r8a.large · tmpfs c8a.large (EBS gp3) and r8a.large (tmpfs) share the EPYC 9R45 (RAM differs: 4 vs 16 GiB): Rust even within run-to-run spread (3m 06s–3m 22s on EBS gp3, 2m 53s–3m 20s on tmpfs).
- c8a.xlarge · EBS gp3 and r8a.xlarge · tmpfs c8a.xlarge (EBS gp3) and r8a.xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 8 vs 32 GiB): Rust even within run-to-run spread (1m 47s–1m 55s on EBS gp3, 1m 47s–1m 58s on tmpfs); TypeScript 49s on tmpfs vs 1m 12s.
- c8a.2xlarge · EBS gp3 and m8a.2xlarge · gp3 1000 MiB/s c8a.2xlarge (EBS gp3) and m8a.2xlarge (gp3 1000 MiB/s) share the EPYC 9R45 (RAM differs: 16 vs 32 GiB): Rust even within run-to-run spread (1m 12s–1m 23s on EBS gp3, 1m 12s–1m 21s on gp3 1000 MiB/s); TypeScript 35s on gp3 1000 MiB/s vs 48s; Docker 4m 39s on gp3 1000 MiB/s vs 7m 49s.
- c8a.2xlarge · EBS gp3 and m8a.2xlarge · tmpfs c8a.2xlarge (EBS gp3) and m8a.2xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 16 vs 32 GiB): Rust even within run-to-run spread (1m 12s–1m 23s on EBS gp3, 1m 13s–1m 18s on tmpfs); TypeScript 32s on tmpfs vs 48s; Docker 3m 32s on tmpfs vs 7m 49s.
- c8a.2xlarge · EBS gp3 and r8a.2xlarge · tmpfs c8a.2xlarge (EBS gp3) and r8a.2xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 16 vs 64 GiB): Rust even within run-to-run spread (1m 12s–1m 23s on EBS gp3, 1m 13s–1m 22s on tmpfs); TypeScript 32s on tmpfs vs 48s; Docker even within run-to-run spread (7m 16s–9m 27s on EBS gp3, 3m 36s–11m 51s on tmpfs).
- m8a.2xlarge · EBS gp3 and r8a.2xlarge · tmpfs m8a.2xlarge (EBS gp3) and r8a.2xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 32 vs 64 GiB): Rust even within run-to-run spread (1m 13s–1m 21s on EBS gp3, 1m 13s–1m 22s on tmpfs); TypeScript 32s on tmpfs vs 45s; Docker even within run-to-run spread (5m 33s–6m 28s on EBS gp3, 3m 36s–11m 51s on tmpfs).
- m8a.2xlarge · gp3 1000 MiB/s and r8a.2xlarge · tmpfs m8a.2xlarge (gp3 1000 MiB/s) and r8a.2xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 32 vs 64 GiB): Rust even within run-to-run spread (1m 12s–1m 21s on gp3 1000 MiB/s, 1m 13s–1m 22s on tmpfs); TypeScript even within run-to-run spread (34s–35s on gp3 1000 MiB/s, 30s–34s on tmpfs); Docker even within run-to-run spread (4m 13s–7m 11s on gp3 1000 MiB/s, 3m 36s–11m 51s on tmpfs).
- m8a.large · EBS gp3 and r8a.large · tmpfs m8a.large (EBS gp3) and r8a.large (tmpfs) share the EPYC 9R45 (RAM differs: 8 vs 16 GiB): Rust even within run-to-run spread (2m 51s–3m 07s on EBS gp3, 2m 53s–3m 20s on tmpfs); TypeScript 1m 26s on tmpfs vs 1m 49s.
- m8a.xlarge · EBS gp3 and r8a.xlarge · tmpfs m8a.xlarge (EBS gp3) and r8a.xlarge (tmpfs) share the EPYC 9R45 (RAM differs: 16 vs 32 GiB): Rust even within run-to-run spread (1m 46s–1m 58s on EBS gp3, 1m 47s–1m 58s on tmpfs); TypeScript 49s on tmpfs vs 1m 03s; Docker 4m 55s on tmpfs vs 9m 08s.
- m9g.xlarge · EBS gp3 and r9g.xlarge · tmpfs m9g.xlarge (EBS gp3) and r9g.xlarge (tmpfs) share the Neoverse-V3 (RAM differs: 16 vs 32 GiB): Rust even within run-to-run spread (1m 47s–1m 52s on EBS gp3, 1m 49s–1m 56s on tmpfs); TypeScript 49s on tmpfs vs 58s; Docker even within run-to-run spread (8m 36s–9m 31s on EBS gp3, 5m 08s–8m 42s on tmpfs).
- m9g.large · EBS gp3 and r9g.large · tmpfs m9g.large (EBS gp3) and r9g.large (tmpfs) share the Neoverse-V3 (RAM differs: 8 vs 16 GiB): Rust even within run-to-run spread (2m 50s–3m 07s on EBS gp3, 2m 55s–5m 01s on tmpfs); TypeScript even within run-to-run spread (1m 39s–2m 01s on EBS gp3, 1m 23s–1m 44s on tmpfs).
- m9gd.xlarge · local NVMe and r9g.xlarge · tmpfs m9gd.xlarge (local NVMe) and r9g.xlarge (tmpfs) share the Neoverse-V3 (RAM differs: 16 vs 32 GiB): Rust even within run-to-run spread (1m 45s–1m 56s on local NVMe, 1m 49s–1m 56s on tmpfs); TypeScript even within run-to-run spread (49s–51s on local NVMe, 48s–55s on tmpfs); Docker even within run-to-run spread (6m 10s–6m 54s on local NVMe, 5m 08s–8m 42s on tmpfs).
- m9gd.large · local NVMe and r9g.large · tmpfs m9gd.large (local NVMe) and r9g.large (tmpfs) share the Neoverse-V3 (RAM differs: 8 vs 16 GiB): Rust even within run-to-run spread (2m 50s–2m 56s on local NVMe, 2m 55s–5m 01s on tmpfs); TypeScript even within run-to-run spread (1m 40s–1m 53s on local NVMe, 1m 23s–1m 44s on tmpfs).
Fixed costs not in cost per build
- Commercial Starter license $350/yearfewer than 50,000 runners a month; one fee per legal entity, every GitHub org and AWS account
- Commercial Growth license $1,050/yearfewer than 200,000 runners a month
- Commercial Scale license $2,100/yearfewer than 500,000 runners a month
- Commercial Enterprise license $4,200/year500,000 or more runners a month; priority support, Slack Connect, source-code access
The license scales with runners launched a month, never with minutes.
Free license for nonprofit, open-source, educational and personal non-commercial projects; 15-day free trial on the Starter tier. EC2 and EBS are paid to AWS, with no markup. Checked 2026-09-30
Head-to-heads RunsOn m8a.xlarge · EBS gp3 vs GitHub 4-coreRunsOn m8a.xlarge · EBS gp3 vs Blacksmith 4 vCPURunsOn m8a.xlarge · EBS gp3 vs Namespace 4x16RunsOn m9g.xlarge · EBS gp3 vs Blacksmith 4 vCPU (arm64)
Over timeone point per run · since Sep 30
Loading every run…