Comment on It's a good thing they don't actually think lol
ignotum@lemmy.world 23 hours agoPlayed around with a tiny 1.5b deepseek model, it was thinking for a loooong while before finally answering my question.
By then it had completely forgotten what the original question was and had instead hallucinated a new question which it then gave me an answer for
I give it a perfect 7/10
DavidGarcia@feddit.nl 54 minutes ago
yeah I wouldn’t bother with anything below 4-7b for this exact reason.