Claude Pricing Explained

Claude API pricing depends on the exact model and the number of input and output tokens processed.

How the estimate is calculated

A useful reference point is Claude Sonnet 4.6. Anthropic lists standard global API pricing starting at $3 per million input tokens and $15 per million output tokens. Anthropic also offers cost-reduction mechanisms such as prompt caching and batch processing.

Suppose an application sends 5 million input tokens and generates 1 million output tokens.

What can change the total

The simplified calculation would be:

Input: 5 × $3 = $15

Plan with a realistic usage example

Output: 1 × $15 = $15

Estimated total: $30

Practical checks before you proceed

This calculation intentionally ignores caching, batch discounts, special modes, tools, and other potential charges.

It is important not to refer to a single number as “the Claude price.” Anthropic offers different models for different workloads, and their prices can vary considerably.

Practical checks before you proceed

For an application processing large volumes of repeated context, prompt caching can be especially important. For asynchronous workloads that do not require immediate responses, batch processing may also provide meaningful savings.

When comparing Claude with GPT or Gemini, compare complete workloads rather than input rates alone. A model that costs more per million tokens could potentially need fewer requests, produce shorter outputs, or deliver better results for a particular task.

Practical checks before you proceed

Last reviewed: August 2026. Check Anthropic's official pricing page before using these numbers for purchasing or financial forecasts.