Cost Cutter
Free archive
Cheaper. After the retries.
Same number of useful results.
Every attempt counted.
How the estimate works
Cost per accepted result = average attempt cost × attempts per task ÷ pass rate. Pass rate measures tasks accepted after all retries. Token averages include rejected attempts; time assumes sequential attempts. Monthly cost targets your requested number of accepted results.
Starting values are examples, not measured prices or savings. Both setups use the same average token counts. Text input and output only; add cache writes, uncounted reasoning, tools, multimodal usage, infrastructure and fees separately. Fallback model costs are not included. Use the kit to measure your own workload.
In the download
- Cost per accepted result
- Compare your own cost, retry, and acceptance assumptions in the browser.
- An offline example
- Inspect synthetic results and a local report. This is not a measured model recommendation.
Details
What runs in my browser?
The calculator uses the assumptions you enter. The offline report contains synthetic example data, not measured model performance.
Can I use it for client work?
The original license in your download governs use of the offline example.
How do I keep my work?
Download the calculator results or use the offline example. There is no account or paid service attached to the free tool.
What happened to the paid kit?
Cost Cutter has been withdrawn from the paid lineup. The calculator and offline sample remain free. For order help: support@rival.tips.