How to Estimate Azure OpenAI API Costs for a 30-Minute Chat

Select Language:

If you’re trying to figure out how much a 30-minute chat with GPT-4 in Azure will cost, here’s a simple way to estimate it.

First, you need to understand how many tokens are used during the conversation. Tokens are like small pieces of words, and the AI’s cost depends on how many tokens it processes. Typically, in a minute, a person talks about 125-150 words, which equals roughly 185-220 tokens. If the AI responds with similar length messages, then for each minute, both user input and AI output together might total around 400-500 tokens.

For a half-hour chat, this adds up to about 12,500-15,000 input tokens and 6,000-7,500 output tokens. Since Azure charges based on the number of tokens processed, you can calculate the cost by taking the rate per million tokens and multiplying it by the number of tokens used in your conversation.

Here’s the basic formula to estimate the cost:

Total Cost = (Rate per million tokens for input × 0.015) + (Rate per million tokens for output × 0.0075)

To get the most accurate estimate, check the latest prices on the official Azure OpenAI pricing page. Keep in mind that if the responses are longer or include additional instructions, your token count might be a little higher, which will affect your costs.

You can also use the Azure Pricing Calculator to customize your estimates based on your actual usage. Just plug in your details and see what the total might be.

Hopefully, this helps you to plan your budget for using GPT-4 in Azure. If you have more specific questions, just check the latest prices or use the calculator for an accurate estimate.