Sub-100ms Time to Interactive
StateSet Sandbox achieves industry-leading TTI by combining warm pod pools, pre-warmed TCP connections, and inline environment injection. Run the benchmark yourself to verify.
TTI Comparison: Sandbox Providers
Time to Interactive = sandbox create + first code execution
Competitor data from ComputeSDK and Agent Arena benchmarks. StateSet value is median of live runs with warm pool.
Live Benchmark
Runs 10 sandbox create + exec + stop cycles on the production cluster
Historical Results
No historical data yet. Run a benchmark to start collecting data.
Methodology
TTI (Time to Interactive) measures the time from a sandbox create request to the completion of the first code execution. It is the most important metric for agent-facing sandbox providers because it determines how quickly an AI agent can start working.
Each benchmark run performs 10 iterations of:
- Create a sandbox (claim from warm pool)
- Execute
echo hello - Stop and cleanup the sandbox
The reported TTI is the median of create_ms + exec_ms across all successful runs. The benchmark runs on the production cluster in us-central1, using the same code path as the public API.
Results are rate-limited (1 request per 60 seconds) and cached for 60 seconds. The controller stores the last 100 results in memory (resets on restart).