Estimate Your Monthly Claude API Spend Before You Scale
Plug in your daily request volume and the typical input and output token sizes per call, and this calculator projects your monthly Claude API bill across Opus 4.8, Sonnet 4.6, and Haiku 4.5 using current published rates.
| Model comparison | Per request | Monthly |
|---|
How the monthly estimate is calculated
Claude's API bills separately for the tokens you send (input) and the tokens the model generates (output), each metered per million tokens. This tool turns a single per-request profile into a full monthly projection so you can size a budget before you flip on production traffic.
The core formula runs per request, then scales out. For one call the cost is:
perRequest = (inputTokens / 1,000,000 × inputRate) + (outputTokens / 1,000,000 × outputRate)
Daily cost multiplies that by your request volume, and the monthly figure multiplies daily by your billing days: monthly = perRequest × requestsPerDay × days. The token counter reports total throughput as (inputTokens + outputTokens) × requestsPerDay × days, which is handy when you are checking projected usage against a rate-limit tier.
The cached input share field models prompt caching, which most cost tools ignore. When a stable prefix (a long system prompt, retrieved documents, a fixed tool list) is served from cache, those input tokens bill at roughly one tenth of the base input rate. The calculator splits your input tokens into a cached slice and a full-price slice: effectiveInput = uncachedShare × inputRate + cachedShare × inputRate × 0.1. Output tokens are never cached, so they always bill at full rate. Pushing the cached share from 0% to 80% on a retrieval-heavy workload visibly drops the bill — that is the single biggest lever most teams have, and the reason it sits in the input form rather than buried in fine print.
The comparison table re-runs the same per-request math against all three models at once, so you can see the Opus-to-Haiku spread for an identical workload. A classification or routing call that is fine on Haiku does not need Opus pricing, and the table makes that tradeoff concrete in dollars rather than abstract per-token rates. Rates reflect Opus 4.8 at $5/$25, Sonnet 4.6 at $3/$15, and Haiku 4.5 at $1/$5 per million input/output tokens.