How to estimate pricing for using OpenAI API

OpenAI’s ChatGPT API enables developers to create interactive and dynamic conversations with the AI model. As developers begin to integrate the API into their applications, it’s crucial to estimate the cost of generating a certain number of words to manage expenses and allocate resources efficiently. Having a clear understanding of the cost structure allows businesses to develop sustainable and cost-effective AI strategies.

In this blog post, we’ll discuss how to estimate the price for generating a specific number of words using the OpenAI API and take a look at an example.

Understanding OpenAI API Pricing

Tokens and Words

OpenAI’s API charges users based on the number of tokens. In the context of the API, a token can be a single character or a word, on the language. On average, 1000 tokens in English correspond to about 750 words. However, the actual number of words may vary depending on the specific content generated.

Cost per Token

To calculate the price for generating a certain number of words, we need to convert the desired number of words into tokens and then multiply by the cost per token.

For your information, here are the prices for different OpenAI base language models:

• GPT-4 Turbo ($0.03/1k tokens)
• GPT-4 Turbo Vision ($0.03/1k tokens)
• GPT-4 8K ($0.06/1k tokens)
• GPT-4 32K ($0.12/1k tokens)
• ChatGPT ($0.002/1k tokens)
• Davinci ($0.02/1k tokens)
• Curie ($0.002/1k tokens)
• Babbage ($0.005/1k tokens)
• Ada ($0.0004/1k tokens)

Estimating Price for Generating a Specific Number of Words

To estimate the price for generating a certain number of words, follow these steps:

1. Determine the desired number of words: Decide on the number of words you want the API to generate.
2. Convert words to tokens: Use the average conversion rate of 1000 tokens which is equal to 750 words to calculate the approximate number of tokens required for your desired word count.
3. Calculate cost: Multiply the number of tokens by the cost per token to obtain the estimated price for generating the desired number of words.

Here’s a simple formula to help with the calculation:

Estimated Price = (Desired Number of Words / 750) * (Cost per 1000 Tokens)

But… OpenAI bills you based on the number of tokens sent in your prompt plus the number of tokens returned by the API. So the modified formula should be:

Estimated Price = (Number of Words Sent as Prompts * Cost per 1000 Tokens / 750) + (Number of Words Generated * Cost per 1000 Tokens / 750)

Keep in mind that the actual number of tokens might vary from the estimate. To get a more accurate cost, you can count the tokens used in the API response and multiply by the cost per token.

Example:

Let’s say we want to generate approximately 1500 words of content using the GPT-4 8K API. Using the formula mentioned earlier, we can estimate the cost as follows:

• Desired number of words: 1500 words
• Completion price: $0.06 per 1000 tokens (or per 750 words)
• Prompt length: 100 words (let’s assume)
• Prompt price: $0.03 per 1000 tokens (or per 750 words)

Estimated price: (1500 / 750 * 0.06) + (100 / 750 * 0.03) = $0.124

Thus, the estimated price for generating 1500 words using the GPT-4 8K API is $0.124. Remember, this is an estimate, and the actual cost may vary depending on the exact number of tokens used in generating the content.

Final words

Estimating the cost of generating content using the OpenAI API is essential to manage resources effectively and build sustainable AI-powered applications. But it doesn’t have to be this complicated…

I have created a simple OpenAI Pricing Calculator, that quickly gives you an estimate. Make sure to check it out.