Free. No account. No telemetry. One executable per platform.
Quickstart
--quick gives a fast first pass in minutes; the full default
suite takes substantially longer because it replays complete inference traces — that depth
is the point.
Cache clearing — required for valid storage numbers — needs elevated
privileges. On Windows: right-click → Run as administrator. On Linux:
sudo ./cai_benchmark_cli --all.
CAI_Benchmark_CLI --all # full suite (storage + memory) CAI_Benchmark_CLI --storage # storage workloads only CAI_Benchmark_CLI --memory # memory workloads only CAI_Benchmark_CLI --quick # faster pass, fewer runs
Three scores on screen; full per-metric detail exported with
--json results.json and --xlsx report.xlsx.
Integrity
Every release ships with a SHA256SUMS.txt on the
GitHub releases page.
# Windows (PowerShell) Get-FileHash .\CAI_Benchmark_GUI.exe -Algorithm SHA256 # Linux sha256sum -c SHA256SUMS.txt