# Cost Curve Workbench

Stage: Preview

Audience: Finance and platform teams

## Summary

A calculator-style tool for converting token pricing into workload cost curves, batch discounts, cache effects, and per-resolution economics.

## Buyer Problem

Token prices do not tell a finance team what production AI will cost. The Workbench converts workload shape into cost per accepted output, including retries, tool calls, cache hit rate, batching, and review fallout.

## Metrics

- Cost per 1k tasks
- Cache savings
- Batch savings
- Reasoning-token exposure

## Deliverables

- Cost curve
- Scenario table
- Savings waterfall
- Budget envelope

## Buyer Questions

- What will this cost at 10k, 100k, or 1M workflows?
- How much do retries and human review change the answer?
- When does prompt caching materially matter?
- Which model class is cheap after quality is included?

## Demo State

Live calculator surface is available; next release should connect to official provider pricing refresh scripts.

Demo readiness: 82/100

Missing for live demo:
- Product walkthrough video
- Client-approved example
- Real run export

## Connected Evidence

- [Cost Efficiency Index](/leaderboards#cost-efficiency-index)
- [Cost curves article](/articles/cost-curves-for-frontier-reasoning-models)
- [Prompt caching article](/articles/prompt-caching-batch-api-and-the-real-ai-cost-stack)

## Visual Preview

![Cost Curve Workbench live Studio preview screenshot](/reports/studio/previews/cost-curve-workbench.png)