Cost & Capacity Planning
How do I split costs between execution and model layers for agents?
Splitting costs between execution and model layers is feasible with targeted separate usage tracking of metrics for each workload component across your full deployment stack. Execution layer costs tie to workflow steps, retries, and allocated resource time, while model costs link to inference calls; tag each component’s usage to map spend accurately across your entire stack. Keep in mind that durable execution does not track model-specific costs on its own.
Was this article helpful?
Your feedback helps improve Diagrid's FAQ experience.
Keep reading
More Diagrid FAQ articles
- Cost & Capacity Planning
What core factors drive higher costs as my agent workload volume scales?
This FAQ explains how scaling agent workloads leads to higher costs through three core factors related to underlying compute and model resources.
- Cost & Capacity Planning
What’s the difference between model and execution layer costs for agent workloads?
Clarify distinct cost categories for model and execution layers in production agent work — model layer costs stem from inference calls, scaling with the
- Cost & Capacity Planning
How do retries and long-running waits affect agent workload spend?
Explain how retries and extended shifts change total spend for agent execution workloads — each retry adds extra execution cycles and additional separate
