Ask HN: What is the least sycophantic frontier LLM?
TL;DR
My daily driver is Gemini, and 3.5 Flash seems more sycophantic and malleable than Gemini Pro 3.1, which is a pretty big deal for me -- I really need as much objectivity and impartiality from the LLM as I can get. So I'm contemplating switching to Claude or ChatGPT, and I wanted to ask about your experiences -- does any frontier model really stand out here? ycombinator. com/item?
Nauti's Take
Promising signal: sycophancy is finally getting mainstream attention — overly agreeable LLMs simply fail at serious research and decisions. Catch: the impression varies a lot across tasks and personas, and models get retuned every few weeks.
For teams using LLMs for analysis or code review, the move is to bake sycophancy probes into the workflow — community vibes alone aren't a selection criterion.
Sources