Claude API Cost Calculator

How Claude pricing works

Anthropic prices Claude per token, with a wide spread across the family. Haiku 3.5 ($0.80 / $4) is the budget workhorse, Sonnet 4 ($3 / $15) is the balanced default, and Opus 4 ($15 / $75) is the premium reasoning model — nearly 20× Haiku's price. Across all three, output is billed about five times the input rate, so the length of Claude's responses is the biggest single factor in your bill.

Two practical moves cut Claude costs the most. First, prompt caching: if you reuse a long system prompt or document context across requests, cached input is billed at a large discount — set the toggle above to see the effect. Second, tier down: a lot of production traffic that defaults to Sonnet or Opus runs perfectly well on Haiku, and the table shows what that switch is worth at your volume. Deciding between providers? Put Claude head-to-head with GPT-4o and Gemini on the AI API cost calculator, or price a whole feature with the AI app cost estimator. Setup and keys are in the Anthropic guide.

FAQ

What's Claude Sonnet's cost per 1M tokens? About $3 input and $15 output (reference, July 2026).

Is Claude cheaper than GPT-4o? Sonnet 4 ($3/$15) is slightly pricier than GPT-4o ($2.50/$10) on output; Haiku undercuts both. It depends on the model tier you pick.

How this calculator works

The Claude API Cost Calculator estimates what you will pay Anthropic to run the Claude API at your expected usage. You pick a Claude model — Sonnet 4, Haiku 3.5, or Opus 4 — and enter your input tokens per request, output tokens per request, and requests per day. The tool applies each model's separate per-token rates for input and output, multiplies by your daily volume, and scales the result into a monthly figure. The main cost drivers are model choice and token counts: Opus costs several times more per token than Sonnet or Haiku, and because output tokens are priced higher than input tokens, long generated responses often dominate the bill more than long prompts do.

The key trade-off is model capability versus price. A larger model like Opus may solve a task in one call, while a cheaper model like Haiku costs a fraction per token but might need retries or longer prompts, so compare total spend rather than the headline rate. The prompt caching input matters when requests share a large, repeated context: cached input tokens bill at a reduced rate, so the same prompt sent thousands of times per day can drop sharply in cost. Adjust tokens per request and daily volume to see which combination of model and caching keeps you within budget before you commit to a design.

Frequently asked questions

How much does the Claude API cost?

Reference pricing (July 2026): Claude Sonnet 4 is about $3 input / $15 output per million tokens, Claude Haiku 3.5 about $0.80 / $4, and Claude Opus 4 about $15 / $75. Output costs roughly 5x input across the lineup, so answer length is the main cost driver.

Which Claude model is cheapest?

Claude Haiku 3.5 is the cheapest, at roughly $0.80 input / $4 output per million tokens — about 4x cheaper than Sonnet and nearly 20x cheaper than Opus. Use Haiku for routine, high-volume tasks and reserve Sonnet or Opus for work that needs more reasoning.

Claude lineup vs alternatives — same workload

How Claude pricing works

FAQ

How this calculator works

Frequently asked questions

How much does the Claude API cost?

Which Claude model is cheapest?