I'd like an initial cost estimate before submitting a prompt. An example use case would be data analysis on a large spreadsheet. I'd like to compare costs for say, Claude Opus vs. Claude Sonnet vs. GPT-4o.

This is a similar idea to the feature titled "Support cost estimation for custom models (e.g., models added via Open Router)" but this initial cost estimation would work for pre-loaded models and OpenRouter models. Like the person in that post mentioned, the cost doesn't have to be exact, it just have to be a somewhat accurate estimate that can be compared across models and prompts. You could approximate the number of tokens and then based on the model pricing you can estimate costs.

Here is a screenshot showing an example.