Claude API Cost Calculator
Estimate your monthly Claude API bill in seconds. Adjust the model, the volume, and the token sizes — see the cost breakdown live. Includes optional prompt-caching savings.
How Claude API pricing works
Anthropic charges per token — roughly a word, or a few characters of non-English text. Two meter readings per call: input tokens (your prompt + system instructions + any cached context) and output tokens (Claude's response).
Pricing is per million tokens, varies by model:
- Claude Haiku 4.5 — cheapest, fastest. Good for high-volume routing and simple tasks.
- Claude Sonnet 4.5 — middle tier. The default for most real work.
- Claude Opus 4.1 — most capable, most expensive. Reserve for hard reasoning tasks.
Prompt caching can cut input costs to ~10% for any prefix you reuse — system prompts, knowledge docs, few-shot examples. Toggle it in the calculator above to see the difference.
Want to use Claude without writing a single API call?
Most people don't need to build their own API integration. Claude.ai with Projects does 90% of the work — and our 4 packs (Skills, Agents, Projects, Studio) give you the shortcut to power-user level without any coding.
- Skills Pack — 2,000+ pre-built Claude skill files
- Agents Pack — 107 specialized Claude agents for Claude Code
- Projects Pack — 50 ready-to-import Claude.ai Projects
- Claude Studio — the 40-lesson interactive course