OpenAI API Pricing Calculator: Cost of ChatGPT API for ERP Workloads

How the API cost is calculated

API pricing is per million tokens, billed separately for input and output. Monthly cost equals requests per month times (input tokens times input price plus output tokens times output price) divided by one million. Output tokens usually cost three to five times more than input tokens, so a short answer to a long prompt is cheaper than the reverse. The cached share setting models prompt caching, where a repeated prefix such as your system prompt and tool definitions is billed at a discount. Set it to zero if you are unsure.

• Tokens are roughly four characters or three quarters of a word in English
• ERP data pasted into the prompt is the biggest input driver
• Reasoning models can bill hidden thinking tokens as output
• Check batch and cached-input discounts on the price page

Reading the model tier comparison

The three tier rows apply the same workload to a small, a standard and a frontier price level, so you can see how much model choice matters compared with volume. Often a small model handles summaries and extraction, while a larger one is reserved for hard reasoning, which is why routing cheap tasks to cheaper models saves most of the bill. The chart sweeps requests per day so you can see where each tier crosses a budget line. The cost per machine figure divides the bill by the machines or users you serve, which is the number finance usually asks for.