Running ChatGPT does not require significant computing power, as the model is hosted on OpenAI’s servers and makes predictions via API calls. This system design enables use in everyday devices like a personal computer or a smartphone. However, training ChatGPT is a highly compute-intensive process which requires multiple high-performance GPUs and could take weeks or months, depending on the dataset size and model architecture. As per OpenAI, training GPT-3, which is a similar model to ChatGPT but larger, took weeks even on a cluster of high-end GPUs.
