sandbox provider

OpenComputer

Composite 48.0 / 100·100 runs per mode·measured 2026-08-14

Rank · sandboxes#20 / 26

Composite48.0 / 100

Full breakdown

median of 100 runs

Every mode ComputeSDK records for OpenComputer, straight from the raw run. TTI is the round-trip to a live sandbox: create a box, run a command, wait for the answer.

Composite

48.0 / 100

latency percentiles × reliability

Median TTI

2.12 s

one create at a time

P95 TTI

7.12 s

95th-percentile cold start

Burst ×100

2.07 s

100 concurrent creates

Staggered

76 ms

100 creates · 200 ms apart

Cost / hr

—

normalized to 1 vCPU + 2 GB

Success rate

88%

runs that completed

How it's measured

ComputeSDK re-runs the sandbox suite every day in public GitHub CI and commits the raw per-iteration results to an open repo. We read those results straight from the source and aggregate the medians shown here — the composite is ComputeSDK's own headline score, folding reliability into the latency percentiles so a degenerate run can't top the board. See the ComputeSDK leaderboard ↗

Source: github.com/computesdk/benchmarks ↗ · aggregated 2026-08-14