I feel like recent models have been consistently getting more sloppy and increasingly claude-ism heavy (load-bearing seams galore) with every new release.
I was hoping this trend would reverse in newer major version releases but I just tried out Opus 5 and it's similarly shit at writing. Fable is slightly better but not by far (and of course ridiculously more expensive). Adding skills helps but only mildly and not much at all for longer prose.
Have you guys used any model lately that you liked for its writing style? I remember being pleasantly surprised by GPT-4.5 but since that's been discontinued, I am on the lookout for a better candidate.
Comments URL: [https://news.ycombinator.com/item?id=49045140](https://news.ycombinator.com/item?id=49045140)
Points: 2
source & further reading
news.ycombinator.com — original article
Companies are optimizing models for specific benchmarks
Show HN: Modelfac
Claude used my pipeline to find a counterexample to the Jacobian conjecture