{"i":14,"slug":"subagent-fan-out-cost","chart":{"id":"fan-out-wall-time","title":"Serial duration scenarios and one-unit reference times (calculation)","subtitle":"Serial = N × median. We did not measure parallel completion time.","kind":"grouped-bar","unit":"seconds","yLabel":"Seconds","whisker":"minmax","series":{"$k":["name","points"],"$r":[["Serial scenario, 10 units",{"$k":["label","value","n","lo","hi"],"$r":[["Claude Code minimal reply (Haiku 4.5; Sonnet 5.5 price)",25.29,5,22.73,33.82],["Claude Code · Claude Sonnet 5.5",23.09,15,21.73,77.25],["Claude Code · Claude Opus 5.5",27.45,15,24.7,89.14],["Claude Code · Claude Haiku 4.5",44.28,15,31.64,235.69],["Claude Code · Claude Fable 5.1",19.44,15,14.09,98.28],["Codex CLI · GPT-6.1 Sol (medium)",56.49,15,40.95,254.61],["Full worker attempt · Crewshal · Sonnet 5.5",5754,33,978,32454],["Codex CLI minimal reply (default model)",59.99,5,53.67,65.06]]}],["Serial scenario, 100 units",{"$k":["label","value","n","lo","hi"],"$r":[["Claude Code minimal reply (Haiku 4.5; Sonnet 5.5 price)",252.9,5,227.3,338.2],["Claude Code · Claude Sonnet 5.5",230.9,15,217.3,772.5],["Claude Code · Claude Opus 5.5",274.5,15,247,891.4],["Claude Code · Claude Haiku 4.5",442.8,15,316.4,2356.9],["Claude Code · Claude Fable 5.1",194.4,15,140.9,982.8],["Codex CLI · GPT-6.1 Sol (medium)",564.9,15,409.5,2546.1],["Full worker attempt · Crewshal · Sonnet 5.5",57540,33,9780,324540],["Codex CLI minimal reply (default model)",599.9,5,536.7,650.6]]}],["Serial scenario, 500 units",{"$k":["label","value","n","lo","hi"],"$r":[["Claude Code minimal reply (Haiku 4.5; Sonnet 5.5 price)",1264.5,5,1136.5,1691],["Claude Code · Claude Sonnet 5.5",1154.5,15,1086.5,3862.5],["Claude Code · Claude Opus 5.5",1372.5,15,1235,4457],["Claude Code · Claude Haiku 4.5",2214,15,1582,11784.5],["Claude Code · Claude Fable 5.1",972,15,704.5,4914],["Codex CLI · GPT-6.1 Sol (medium)",2824.5,15,2047.5,12730.5],["Full worker attempt · Crewshal · Sonnet 5.5",287700,33,48900,1622700],["Codex CLI minimal reply (default model)",2999.5,5,2683.5,3253]]}],["One-unit median reference (not a parallel bound)",{"$k":["label","value","n","lo","hi"],"$r":[["Claude Code minimal reply (Haiku 4.5; Sonnet 5.5 price)",2.529,5,2.273,3.382],["Claude Code · Claude Sonnet 5.5",2.309,15,2.173,7.725],["Claude Code · Claude Opus 5.5",2.745,15,2.47,8.914],["Claude Code · Claude Haiku 4.5",4.428,15,3.164,23.569],["Claude Code · Claude Fable 5.1",1.944,15,1.409,9.828],["Codex CLI · GPT-6.1 Sol (medium)",5.649,15,4.095,25.461],["Full worker attempt · Crewshal · Sonnet 5.5",575.4,33,97.8,3245.4],["Codex CLI minimal reply (default model)",5.999,5,5.367,6.506]]}]]},"note":"A serial total is a sum of durations. N × the median is a scenario, not its measured total or expected value. Whiskers show observed per-unit ranges, scaled by N for serial scenarios. They are not confidence intervals. Simultaneous completion depends on the maximum duration, not the median. The slowest recorded worker attempt took 54.1 min. We did not measure rate limits, machine limits or contention.","sourceIds":["calc-fan-out","agent-routing-overhead","agent-provider-h2h","agent-swebench-c1","agent-swebench-c2"]}}