The 2020 GPT-3 paper describes GPT-3 as an autoregressive language model with 175 billion parameters, 10 times more than any previous non-sparse language model.

Checked against its source on October 17, 2026

In the source's own words

“an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model”

Language Models are Few-Shot Learners, Tom B. Brown et al., Advances in Neural Information Processing Systems 33 (NeurIPS 2020); arXiv preprint, 2020, Abstract.

Open the original

What it measures: The abstract reports that scaling up a language model greatly improved task-agnostic, few-shot performance. The model was tested with few examples in the prompt and no gradient updates.
What it does not show: A 2020 model, and a parameter count only. Parameter count does not measure accuracy, and newer models are often larger or trained differently.

Source

Language Models are Few-Shot LearnersPeer-reviewed research · Tom B. Brown et al., Advances in Neural Information Processing Systems 33 (NeurIPS 2020); arXiv preprint, 2020

Cited on

Large language model (LLM)Glossary