Cost per task (agent unit economics)
What one finished agent task really costs: tokens, retries, tool calls and the human time to review it, compared with doing it another way.
Price per million tokens says little on its own. A cheaper model that needs three attempts and a longer review can cost more per finished task than an expensive one that gets it right once. Measure tasks completed, not calls: total spend and review minutes divided by tasks that passed your checks. That number lets you compare models, harnesses and routing rules, and set a sensible cost cap.
Related terms
- Cost cap: A hard spending limit enforced by your code, per user, per day or per month.
- Tokens: The pieces of text models read and write, and the unit you pay for.
- Model routing and fallback: Sending each request to the model that fits it, and to another one when the first fails.
- Trajectory evals: Tests that grade an agent’s whole path to an answer: which tools it called, in what order, how many steps and at what cost, not only the final output.
