This model was developed by researchers at OpenAI to help us understand how the capabilities of language model capabilities scale as a function of the size of the models (by parameter count) combined with very large internet-scale datasets (WebText).
