300 billion parameters or GTFO, eh? There is tons of value to be had from smaller models. Even some state of the art results can be obtained on a relatively sm

300 billion parameters or GTFO, eh? There is tons of value to be had from smaller models. Even some state of the art results can be obtained on a relatively small set of commodity GPUs. Not everything is GPT-scale.