Hacker News new | ask | show | jobs
by jonaustin 98 days ago
Try with qwen 3.5 122b; it has more parameters so a larger corpus of knowledge to draw from than 35b.