Comment by bitmasher9

I often have the expensive models give relatively simple inaccurate answers, even when they cite sources that directly contradict them. The error rate is lower, but you can’t have confidence with llm answers.