Stability AI unveils new FreeWilly language models trained using minimal -- and highly synthetic -- data

The FreeWilly1 and FreeWilly2 LLMs from Stability AI were trained with 600,000 data points -- 10% of the size of the original Orca dataset.