Skip to main content
The Cost evaluator calculates the cost of each prompt response based on actual token usage returned by the LLM provider and its public pricing. Use it to monitor spending, enforce budget thresholds, and optimize prompts for cost efficiency. Adaline automatically calculates costs using the token usage reported by the AI provider. Internally, Adaline maintains the public pricing for each model and uses that to compute the cost per response.
Cost evaluators are not supported for some AI providers (such as OpenRouter, TogetherAI, Custom, etc.) where available models are set by the user.
Want to use custom pricing for your models? Reach out to us at support@adaline.ai to set up a private preview.

​
Set up the Cost evaluator

1

Open Evaluate from your prompt

Open the prompt you want to test and click Evaluate in the prompt header.Opening the Evaluate tab from a prompt
2

Select the evaluator

Add the Cost evaluator from the evaluator menu.Selecting the Cost evaluator
3

Configure the threshold

Give the evaluator a name, link a dataset, and set the cost threshold that determines pass/fail.Configuring the Cost evaluator thresholdChoose from the following threshold operators:
OperatorBehavior
less thanThe response passes if its cost is below your threshold. Use this to enforce budget caps.
greater thanThe response passes if its cost exceeds your threshold. Use this to flag unusually cheap (potentially incomplete) responses.
equal toThe response passes if its cost matches the exact threshold.
4

Run the evaluation

Click Evaluate to execute the evaluation and see the results.Cost evaluator results
Prompt chaining: When your prompt uses prompt variables (child prompts), the Cost evaluator reports the cumulative cost of the entire prompt chain — including the parent prompt and all child prompts executed during the evaluation.

​
When to use

  • Budget enforcement — Set a maximum cost per response to prevent runaway spending.
  • Model comparison — Compare cost efficiency across different models handling the same test cases.
  • Prompt optimization — Identify prompts that are more expensive than expected and refine them to reduce token usage.
  • Cost forecasting — Use evaluation results to estimate production costs at scale.

​
Next steps

Latency Evaluator

Measure response time alongside cost.

Response Length

Control output size to manage costs.