cyberneticlibrary

Benchmark code across 8 concurrent workers

bench-s2-parallelworkflowsetup L31
ch-courtesy/claude-plugins
What it does

Benchmark Sonnet parallel performance and report latency

Best for

Model selection when parallel-agent latency and throughput baselines must be measured.

Inputs
  • · [object Object]
Outputs
  • · [object Object]
Preconditions

Sonnet model available; batch size > 0

Failure modes

Outlier latencies; insufficient samples; throughput underestimated

Trust signals
  • · Per-iteration latency tracking
  • · Throughput calculation
  • · Outlier filtering