I have some confusion regarding HuggingFace’s pricing model, and how it might apply to AI models like Pegasus and GPT.
Here is the pricing model for OpenAI:
They charge from $0.0008 to $0.06 per 1K tokens depending on the model (I think a word is about 1.25 tokens). Does anyone know if the following would be an appropriate comparison between GPT versions, and OpenAI’s models?:
GPT Neo 125M - Ada
GPT Neo 1.3B - Babbage
GPT Neo 2.7B - Curie
GPT-J 6B - Davinci
For paraphrasing you need to pass the original content as input, so assuming an article is a thousand words, HuggingFace would cost $50 for 1K articles or $0.05 per article. Is my math correct there? Or, do you get charged for both the input article, and the output article - so if you paraphrase a 1K word article, that’s 2K words, and so $0.10?
Also how does it work for text generation libraries like GPT? Let’s say you input a 10 word phrase and want 300 words as output, how does HuggingFace’s pricing model apply to that?
Any clarification would be appreciated.