Skip to content

Latest commit

 

History

History
49 lines (29 loc) · 1.61 KB

token-estimation.md

File metadata and controls

49 lines (29 loc) · 1.61 KB

Token estimation

Prompt

Very simple example prompt, assuming prior extraction of text form PDFs:


You will receive a resume of a person in the format of extracted text.
You need to transform this resume into a format called 'Story' that eliminates the chance of a decision based on certain factors. The factors to be eliminated will be described below.
Additionally, the 'Story' needs to be in a human readable format describing S.M.A.R.T goals that apply for this person.
A 'Story' should also contain around 1000 words as an introduction to the person so that a recruiter can quickly gain an overview of a persons experience, skills.

Now I will give you factors that should not be identifyable in the story to make discrimination impossible. These factors are:
- race
- sex
- gender
- sexual orientation
- health
- attractivity


The resume in the form of extracted text from a pdf, follows below:

--- an additional 700 Tokens, from extracted text from a sample resume pdf ---

Total tokens: Around 850

Response

Token number depends on the number of words desired in the response. We assume a story should contain around 400 Words, leading to around 1400 response tokens.

Pricing

Based on https://web.archive.org/web/20240929131353/https://openai.com/api/pricing/?__cf_chl_rt_tk=NVe853hvq27cBWiUvHHcQyjrYt6rMFrZBoo0Hz37WpE-1727615633-0.0.1.1-6100

With the latest model, GPT-4o, the cost would be at minimum: 0.025 $ / prompt & answer. With the older, cheapest, model GPT-4o mini, the cost would be at minimum 0.001 $ / prompt & answer.

Leading to assumption: A24