OpenAI API Pricing Calculator

It depends on the model. As of 2026, GPT-4o is roughly $2.50 per million input tokens and $10 per million output tokens; GPT-4o mini is around $0.15 per million input and $0.60 per million output. That makes GPT-4o mini ~16-20x cheaper for the same volume. This calculator uses the latest published list prices.

A token is roughly a piece of a word — about 0.75 English words on average. The word 'extraordinary' is 3 tokens; 'cat' is 1; punctuation usually gets its own token. To count tokens precisely, use OpenAI's tiktoken library; for budgeting, multiply word count by 1.33 to get a rough token count.

Each turn = (system prompt + conversation history + new user message) input tokens + the response output tokens. Conversations grow as history accumulates, so by turn 5 the input cost can dwarf the output cost. A typical RAG chatbot turn runs 1k-3k input tokens (system + retrieved docs) and 200-500 output tokens — about $0.003-$0.010 on GPT-4o.

For most customer-facing chatbot use cases, yes — it's noticeably weaker than GPT-4o at multi-step reasoning and code, but for grounded Q&A over your documents it's often indistinguishable in user testing. Start with GPT-4o mini, measure quality, and upgrade only the specific calls where you can prove GPT-4o is worth the price.

We update the catalog whenever OpenAI publishes a new pricing page. For real-time confirmation, check openai.com/api/pricing — and remember that enterprise customers with annual commits typically get 10-25% off list, so build that into your forecast if you'll qualify.

Currently chat completion pricing only. For DALL-E, Whisper, or fine-tuning costs, check the OpenAI pricing page directly — those models have different per-unit pricing (per image, per audio minute, per training token) that doesn't fit the same calculator.