3 ms·Mistral-7B does this reliably with no special finetuning, while being 1000x smaller than GPT-4.by MacsHeadroom 3y agoMistral-7B does this reliably with no special finetuning, while being 1000x smaller than GPT-4.