Key Stats About OpenAI’s GPT-3

Photo of author
Written By Thomas Smith

The New York Times called Thomas Smith a "veteran programmer." For over a decade, Smith has led Gado Images, an AI-driven visual content company.

GPT-3 from OpenAI is one of the most powerful generative content systems in the world. Here are some key stats about GPT-3.

  • GPT-3 was trained on nearly 1 trillion words of text (Source)
  • GPT-3 uses 175 billion parameters. Parameters are the parts of a neural network that let it process data. (Source)
  • 82% of GPT-3’s training data comes from websites (via Common Crawl and WebText2), 16% comes from books, and 3% comes from Wikipedia. (Source)
  • Humans identify GPT-3’s output as AI-generated only 52% of the time, which is only slightly better than random chance. (Source)
  • GPT-3 can produce up to 50,000 characters at a time (Source)
  • OpenAI charges about $0.0000266 per word generated by GPT-3’s most popular model. (Source)
  • It costs about 2.6 cents to have GPT-3 generate a 1,000-word article. (Source)

You are welcome to cite our GPT-3 statistics in your own article or paper. Please consider linking to this page if you do so.

The URL is:

Leave a Comment

Receive the latest articles in your inbox

Sign up for Gado Images' free newsletter

If you make a purchase from a link on our site, we may receive a commission at no cost to you, which helps to support our reporting. Thanks!