LLM Cost per Request vs Cost per Accepted Task: A Team Budget Guide

TL;DR: Calculate LLM cost per request from tokens, then budget your product around cost per accepted task. A request is a model attempt; a task is the outcome your customer actually needs. A draft, a checker and a retry can all belong to one task. For a plain request without caching or special pricing conditions, estimated cost is input tokens multiplied by the input rate plus output tokens multiplied by the output rate, with rates expressed per token. When using rates per million tokens, divide both terms by one million. To evaluate a route, add every attempt and divide by the number of outputs that pass the same acceptance criteria. Test Haiku, Sonnet, Opus and Fable against your own examples rather than treating model names as an automatic quality ladder. cocodot bills usage by tokens, with live model rates on its pricing page. New users must complete identity verification before API calls, and trial credit covers selected budget models only. The following worksheet is an application-side evaluation method for teams planning repeated content, support or agent workloads, not a promise of a particular saving.

function plainRequestCost(inputTokens, outputTokens, inputPerMillion, outputPerMillion) {
  return (inputTokens * inputPerMillion + outputTokens * outputPerMillion) / 1_000_000;
}
function costPerAcceptedTask(allAttemptCharges, acceptedTasks) {
  if (acceptedTasks === 0) return null;
  return allAttemptCharges.reduce((sum, charge) => sum + charge, 0) / acceptedTasks;
}