Hacker News new | ask | show | jobs
by MacsHeadroom 931 days ago
Mistral-7B does this reliably with no special finetuning, while being 1000x smaller than GPT-4.