Net monthly savings from routing easy requests to Haiku, after retries and router overhead, computed in your browser.
Routed requests pay Haiku rates. A share of those fail the quality bar and are retried on Sonnet, paying for both calls. The router itself adds a few classifier tokens to every request. The break-even retry rate is where routing stops saving money entirely.
When a large share of requests are simple enough that Haiku's quality is acceptable. At list prices Haiku 4.5 costs about one fifth of Sonnet per token, so routing even 50 percent of traffic roughly halves spend, before retries.
Each retried request pays twice: once on Haiku and once on Sonnet. The calculator includes a retry rate so you can see the net saving, not just the gross discount.
A small-model router or a rules-based classifier both work. The overhead field covers the extra tokens either way; keep it small, a few hundred tokens per request is typical.
I build and ship the ClaudFlow workflow builder, including its eight block types and JSON export format. The agent loop patterns these calculators price are patterns implemented in shipped code, not invented for a blog post. Everything runs client side by design: nothing you type leaves this page. More on the about page.
ClaudFlow is a free, privacy-first suite of AI workflow tools. No accounts, no tracking, no data leaves your browser. Part of the Zovo Tools network by Michael Lip. Token counts are estimates, and actual usage depends on encoding, model version, and prompt structure.