OpenAI Pricing API Guide

When building software around OpenAI, developers often want to calculate costs automatically rather than manually updating formulas throughout the codebase.

How the estimate is calculated

Your application should treat model pricing as configuration.

For example:

What can change the total

`javascript const pricing = { inputPerMillion: 1.25, outputPerMillion: 10 };

const inputCost = (inputTokens / 1_000_000) * pricing.inputPerMillion;

Plan with a realistic usage example

const outputCost = (outputTokens / 1_000_000) * pricing.outputPerMillion;

const total = inputCost + outputCost; `

Practical checks before you proceed

Those example rates correspond to the original gpt-5 standard API rates currently documented by OpenAI.

However, production code should not assume that every GPT model uses those values.

Practical checks before you proceed

OpenAI pricing varies by model and can distinguish standard input, cached input, output, long-context usage, processing modes, and tool-specific charges.

Therefore, store pricing by exact model ID.

Practical checks before you proceed

A cost record might contain:

`text model input_rate cached_input_rate output_rate effective_date `

Practical checks before you proceed

When requests finish, record the actual usage information returned by the API and calculate costs from those values.

This gives you useful metrics such as cost per endpoint, user, organization, feature, and successful workflow.

Practical checks before you proceed

If you're building a public OpenAI cost calculator, also display when your pricing data was last updated.

Avoid silently presenting old prices as current ones. OpenAI's product lineup changes frequently, and financial calculations should always be checked against the official pricing documentation.