Free tool · runs in your browser · no signup
A retry rate is a cost multiplier.
Set your runs per month, your average cost per task, and your retry rate, and see the blended cost per shipped task, plus exactly how much of your spend is the retry tail. Everything runs in your browser, nothing is sent anywhere.
80%
Blended cost per shipped task
$27.5
Claude Sonnet 5 at $5.50/task, 730 runs/month, 80% retries.
One month, itemized
Total API list cost
$4,015
Runs that ship
146
Retry tail (80% of spend)
$3,212
Before you quote this number
- Per-model cost is a relative multiplier against Claude Sonnet 5, verified against Anthropic model pricing card, per-task relative cost index on 2026-09-05.
- Estimates only, at list price. Measure your own average cost per task before trusting any calculator, including this one.
FAQ
Agent costs, straight answers.
The blended cost per shipped task for a month of agent runs: total API list spend across every run, divided by only the runs that ship after retries are backed out. A retry rate does not just add cost, it shrinks the denominator too, which is why the blended number moves faster than the retry percentage suggests.
Each model's baseline cost per task is scaled from a relative-cost index versus Claude Sonnet 5, verified against Anthropic model pricing card, per-task relative cost index on 2026-09-05. Pick a model and the average cost per task you enter is scaled by that model's multiplier before the blend is computed.
Retries are spend on runs that never ship, so they inflate the cost of the runs that do. This calculator itemizes the retry tail as its own dollar figure and as a percent of total spend, so it is visible instead of buried in an average.