Start with the cost formula
OpenAI API pricing is usually calculated from input tokens and output tokens. Input tokens are the prompt, system instructions, retrieved context and conversation history sent to the model. Output tokens are the generated answer.
The practical formula is: input tokens divided by 1,000,000 multiplied by the model input price, plus output tokens divided by 1,000,000 multiplied by the model output price. Monthly cost is that request cost multiplied by expected monthly requests.
Use realistic model rates
For text workloads, OpenAI lists prices per 1M tokens. As of July 22, 2026, examples on the official pricing page include ChatGPT chat-latest at $5 input and $30 output per 1M tokens, and gpt-5.3-codex at $1.75 input and $14 output per 1M tokens. Tool usage, realtime, image, audio and batch pricing may follow separate rules.
That means a product estimate should not use a single flat per-message number. A support bot with short answers can be cheap, while a workflow that sends long documents and asks for detailed analysis can spend much more on output tokens.
| Scenario | Input tokens | Output tokens | Rate example | Estimated request cost |
|---|---|---|---|---|
| Short assistant reply | 1,000 | 300 | $5 in / $30 out | $0.014 |
| Detailed analysis | 8,000 | 1,500 | $5 in / $30 out | $0.085 |
| Long document summary | 40,000 | 2,500 | $5 in / $30 out | $0.275 |
Estimate before implementation
Before building, collect three samples: a short request, a typical request and a worst-case long request. Estimate cost for all three, then multiply by expected monthly traffic. This exposes whether the product needs prompt trimming, caching, smaller models or batching.
For app planning, connect the estimate to product behavior. A chat app should model conversation history growth. A summarization product should model transcript length. A retrieval app should include retrieved chunks, not only the user question.
Use the calculator for validation
The OpenAI calculator on this site lets you paste representative content, estimate tokens locally and compare monthly cost against other LLM providers. Use it as a first-pass planning tool, then confirm final production pricing against your provider invoice and selected API features.
Sources
Pricing changes over time. These examples use official provider pricing pages checked on July 22, 2026.
OpenAI Pricing FAQ
What affects OpenAI API cost the most?
The biggest drivers are input context length, generated output length, model choice, monthly request volume and use of tools such as search or file retrieval.
Is cost per message a reliable estimate?
Only if message size is stable. For real products, estimate tokens from representative examples because conversation history and retrieved context can increase cost quickly.
Should I use the cheapest model?
Use the cheapest model that meets quality requirements. Many production apps route simple tasks to lower-cost models and reserve stronger models for complex cases.