>>87142539These fucking things muck with their own output so much to be 'user-friendly' nowadays and lobotomize themselves it's ridiculous
Until they get so anyone can run a fultext model on their own home GPU open source, even if it takes an hour to get an answer, people are always going to be wary