Yeah, i noticed it too. Something feels off with the reasoning on complex multi-step tasks compared to a few months ago. hard to tell if it's actual regression or just expectations creeping up as you use it more.

Been mixing Opus with Sonnet depending on the task. Sonnet handles most things well enough and Opus for anything that genuinely needs deeper reasoning. Try it out, may be you find it useful

Isn't it the expected thing that LLMs degrade over time?

Not using opus 4.6 but I have heard the same things.

https://old.reddit.com/r/LocalLLaMA/comments/1sgd7fp/its_ins...

There was also another post about how the perceived qualities of these models is going insanely down, something not reflected in benchmarks

I feel like it might be because the costs of GPU is reflecting back up and they might be having a more diluted model which makes it more dumb while still taking the 100$

I personally feel like this theory of these models slowing going down in intelligence until a new model which isn't bogged down intentionally might be of more interest than people think because my experience with even claude sonnet 3.7 when it had first launched was genuinely fascinating and gemini 3.1 premium and it really aligns with my personal experience tinkering with these models.

The AI industry feels quite scam-my to be honest and we would all be forced by IPO or index funds bending backwards to be left holding the bags :-/

It really feels like a great deception being played against the masses.