wzlxp.top builder tools Claude calculator
Estimate Claude API cost from usage tokens or planning text. Cache writes and cache reads are separated so repeated prompts do not hide inside one blended number.
High confidence: based on Claude usage or token-counting API numbers.
Pricing checked 2026-05-09. Claude Sonnet 4.6: input $3/1M, cache write $3.75/1M or $6/1M, cache read $0.3/1M, output $15/1M tokens.
Cost insight: the largest cost driver is input tokens. Cutting fresh input tokens by 50% would save about $90.00 per month.
Estimated planning result only. Prices can change, and provider bills may include taxes, minimums, feature-specific charges, or usage adjustments. Verify production spend in the official provider dashboard.
How to use
Select the model version that matches your planned Messages API workload.
Put repeated prompt context into cache writes and cache reads so recurring prompts do not hide inside one input number.
Use the monthly projection for planning, then verify production counts with Anthropic usage data or token counting.
Claude cost guide
Claude workloads often depend on long reusable context. Keeping cache writes, cache reads, normal input, and output visible makes the estimate easier to trust.
User messages and uncached context are estimated with the model's input token price.
Prompt caching can charge differently when context is first written versus when that cached context is reused.
Claude responses use output token pricing. Message Batches are useful for async jobs where lower cost matters more than immediate response time.
Estimate a reusable instruction block separately from new user messages to see when caching starts paying off.
Model a stable policy document or knowledge base context as cached content, then add each user question as normal input.
Use batch estimates for delayed classification, extraction, or review tasks that can run outside the user request path.
References
The official Token Counting API accepts the same message shape as Claude Messages and returns input token estimates before sending a request. This page does not call that API or collect API keys.
Web search, code execution, and managed-agent session fees are excluded from this first version so the core token and prompt-cache flow stays easy to use.
FAQ
It estimates input token cost, prompt cache write cost, cache read cost, output token cost, Message Batches discounts, and projected daily, monthly, and annual spend.
Claude prompt caching can price cache writes and cache reads differently, so separating them makes repeated prompt costs easier to understand.
No. It runs in the browser for planning. Use Anthropic usage data or the Token Counting API when you need exact production counts.