OPT-175B is Comparable to GPT-3 While Requiring Only 1/7th the Carbon Footprint

OPT introduces a suite of open-access language models from 125M to 175B parameters, replicating GPT-3 performance while using 1/7th the carbon footprint. Researchers gain full transparency via model weights, code, and training logs, enabling reproducible experiments. OPT-175B performs well in NLP and dialogue tasks but retains known LLM limitations, highlighting responsible AI considerations.